Sound Generation and Propagation in Ducts
Sound Generation and Propagation in Ducts
Sound Generation and Propagation in Ducts
Fabrice Ducret
Licentiate Thesis
Stockholm 2006
The Marcus Wallenberg Laboratory for Sound and Vibration Research (MWL)
Department of Aeronautical and Vehicle Engineering
The Royal Institute of Technology (KTH)
1
Postal address
Visiting address
Telephone/E-mail/Web-page
Teknikringen 8
+46 8 7907903
MWL
Stockholm
fabrice_ducret@hotmail.com
SE-100 44 Stockholm
Sweden
http://www.ave.kth.se/
Abstract
This thesis contains three papers investigating problems of interest for noise control in
ducts.
The first part of this thesis treats the sound propagation in rectangular ducts with
flexible walls. Various experimental techniques are performed to measure the internal
sound propagation and radiation to the surrounding. An analytical model is derived to
calculate the coupled propagation wavenumber and radiated sound power. The two-port
formalism is used.
The second part starts with the sound propagation in open ended circular straight pipe
with airflow (a tailpipe). Various aspects such as: acoustic damping, reflection and
transmission at the open termination are investigated. Sound absorption due to vorticity
shed at the opening is also treated. The geometry of the opening is then modified
(oblique cuts, diffusers) and comparisons with the reference straight pipe is made for the
sound transmission and flow induced noise generation. The effect of an upstream bend
close to the opening is also investigated.
In the third part the acoustic impedance of perforated plates are investigated. In
particular the application to small perforation ratios ( 1% ) and holes or slits with
duct, pipe, reflection coefficient, impedance, flow noise, vorticity, damping, scaling
laws, impedance, transmission loss, perforated plate, slit, non linearity.
Content
Paper I
The author gratefully acknowledges the financial support from the following sources:
Paper I has been funded by the Rdet fr ArbetsLivsForskning project (dnr 1999-0082).
Paper II has been funded by the European commission under the Artemis project
(contract number: G3RD-CT-2001-00511).
Paper III has been funded by the European commission under the EDSVS programme
in partnership with Acoustic Technology, rsted, DTU, Lyngby, Denmark.
Related conference papers
Low frequency sound propagation in rectangular ducts, Tenth International Congress
Acknowledgments
For his willingness to teach me and share his acoustic knowledge, I am grateful to Mats
bom.
I am indebted to Susann Boij for the numerous enlightening discussions on the
dimensionless but certainly not pointless Strouhal number. She has provided invaluable
guidance on the writing of this document. Any errors that remain are, of course, my
own. Ragnar Glav has also to be thanked for his time spent on the interpretation of the
results concerning the modelling of the exhaust lines.
I have greatly benefited from the help given to me by Hans Bodn in various areas.
My stay in Denmark could not have been possible without the support of Finn Jacobsen,
Acoustic Technology, rsted, DTU, Lyngby, Denmark.
Down in the lab, technical and moral assistance after rig blowouts has been kindly
provided by Kent Lindgren and Danilo Prelevic.
I have immensely appreciated my friendship with Daniel Backstrm. I have listened
with great interest to his Gaussians theory of life. , ,
. ,
, .
I would like to thank the staff members at MWL for encouragements and support.
Acoustic poetry
if you would blow on my heart, near the sea, weeping,
it would sound with a dark noise,
with the sound of sleepy train wheels,
like wavering waters,
like a leafy autumn,
like blood,
with a noise of moist flames burning the sky,
sounding like dreams or branches or rains,
or foghorns in a dreary port
Extract from Bacarole, Pablo Neruda (1904-1973)
The type of noise you certainly do not want to attenuate...
Table of contents
1
Introduction .............................................................................................................. 1
2.1.1
2.2
2.2.1
2.3
Paper III. Linear and non linear acoustic regimes of perforated plates.......... 25
2.3.1
Main contributions.................................................................................................. 31
References .............................................................................................................. 33
1 Introduction
Airflow is required to expel exhaust fumes and particulates formed during the
combustion process in an internal combustion engine. In a ventilation system, air blown
by fans is required to transport heat to ensure comfort. In addition, noise produced by
elements such as fans, engines or by the flow itself when flow separation occurs is
borne by the flow. Therefore, two fields (acoustic and flow) coexist. Aeroacoustics is
the branch that investigates interaction between these two fields.
Paper I in this thesis deals with the sound propagation in ventilation systems. The flow
speed is of magnitude less than 10 m/s. Thus, phenomena associated with flow is not
investigated in this part. Influence of the flow is considered in paper II where exhaust
tailpipes and various terminations is analysed. For this case flow effects on the sound
propagation and on the sound generation are studied. The main application of this
investigation concerns the automotive industry. No flow investigation has been
performed in paper III where perforated plates are studied. However, flow effects can
play a role to understand the acoustical behaviour of such systems.
In an internal combustion engine, pulsating exhaust gas flow excites the structure.
Engine elements such as the piston generate structural vibrations. This energy is
transmitted through the other components and radiates as sound. In a ventilation system,
the ducted elements are sometimes made of thin flexible aluminium sheets. These
panels are easily excited either by vibrations or by the sound and flow fields present in
the airway and transmit structural vibrations. Vibrating structures are also found in
shells of silencers. Vibroacoustics deals with the types and the behaviour of the
different structural waves that can exist. In paper II and paper III the pipe wall and
perforated samples are considered rigid and therefore structural vibration is neglected.
Paper I includes vibroacoustic effects as the rectangular ventilation ducts studied are
allowed to vibrate and radiate acoustic energy.
Thermal effects and thermoacoustics are not considered in this thesis. Temperature
converters. But for the tailpipes investigated here, the effect of temperature gradients is
not important. The tests reported in paper III have therefore been done at cold
conditions (ambient temperature 20o C ).
This thesis focuses on three acoustic elements found in buildings and cars or trucks,
namely, a duct ventilation system (Paper I), an exhaust system (Paper II) and a
perforated element (Paper III). Noise in fluid machinery systems can be reduced directly
at the source or along the propagation parts which is studied here. The walls of a, e.g.,
ventilation system, can be stiffened to shift the resonance frequency and reduce the
radiating sound. Porous materials can be applied on the walls to absorb outgoing sound
waves. An internal additional sound treatment is often required. The easiest consists of
placing sound absorbing material within the ventilation duct. However, sufficient sound
reduction requires the use of bulky material that takes up space. Thin perforated panels
are often preferred. They are placed in front of walls and form panel absorbers. At
specific frequencies depending on the geometries of the perforations and depth of the
backing cavity, resonance occurs and an enhanced reduction of noise is obtained.
Absorbing sheets and materials can be placed in the cavity to improve the sound
reduction level. Another interesting approach is to reduce the size of the perforations to
the sub-millimetre range, creating so-called micro-perforated panels. For minute
dimensions, it has been shown that acoustic losses occur within the perforations. The
other advantage of sound reduction through micro-perforated elements is the low
frequency efficiency and large bandwidth.
Ducts with thin flexible vibrating walls are used in buildings for airflow transportation
between the various locations for ensuring better living conditions. Ventilation systems
are also operated in many other applications, e.g., cars and trains. The manufacturing
process usually consists of folding thin aluminium or steel sheets into rectangular ducts.
Although, the main purpose is the transportation of a fluid such as air, noise is also
carried out along the duct. Various sources can be thought of, the most obvious one is
the fan or fluid machine connected to the duct. Other sources are flow induced noise due
to flow separation, speech and other external sources which enters the duct via openings
or via breakin through duct walls. The sound transmission to the external environment is
sometimes referred as the breakout effect. This is a low frequency process and promotes
the extraction of the internal acoustic energy. This is of interest as a sound reduction
effect as long as the external environment is not populated. Another effect called
structural flanking transmission takes place when the acoustic energy is transmitted to
the structure and is carried away until it radiates into the core of the duct. Radiation
bypass occurs when the noise is first transmitted out of the duct, propagates in the
external environment and re-enters the duct. Porous materials are lined onto the walls to
maximise the sound absorption and bars or corrugations are applied to stiffen the
structure. The acoustic effects (external radiation to the outside, transmission along and
internal absorption) need to be quantified for the case when the flexible duct is part of
an entire system composed of other duct elements. An analytical approach is preferred
for its simplicity and ease of implementation for calculations involving multiple
acoustic elements. For the low frequency plane wave range, which is of main interest
here, a building block technique where the duct elements are represented as acoustic
two-ports can be applied.
Transmission of sound through walls of flexible ducts has been exhaustively treated by
Cummings (1978). An accurate calculation of the radiation from the complete
rectangular structure requires the solving of the equations of the motion of the walls.
The structural wave theory is used to predict the wall admittance and attenuation in the
wall. Cummings (1978) derives a simple low frequency model based on coupled
acoustic/structure wave system. Cummings (1978) treats the radiating duct as a finite
length source with a single travelling wave. Cummings (1978) obtains good comparison
with experimental results up to 2/3 of the cut-on frequency of the lowest cross-mode of
the equivalent rigid walled. An extreme test of the theory is made (Cummings (1978))
on a short duct with rigid termination and gives good prediction using the single wave
model.
A high frequency approximate asymptotic model has been derived by Cummings (1983)
based on the mass law wall impedance and including a single higher order mode. The
duct is treated as a radiating circular cylinder with the same surface velocity
distribution. The predicted radiated power is slightly underestimated. External lagging
consisting of wrapping up the duct with porous material such as mineral wool and glass
fibre covered by an impervious layer such as plaster or sheet metal has been treated
theoretically by Cummings (1979). External lagging is used for two purposes: thermal
insulation and sound reduction. A study of the stiffness control of the duct walls has
been performed by Cummings (1981) by using materials (Firmalite, polystyrene,
honeycomb) with different stiffness/mass ratios to reduce low frequency acoustic
breakout related to transverse resonances and to increase the fundamental transverse
resonance frequency of the walls. A low frequency approximate method is given to
calculate the transmission loss of ducts. Flanking transmission effects (structural and
breakout/breakin) have been modelled by Cummings (1995) using the reciprocity
theorem. A more accurate and refined modelling has been performed using a finite
element scheme (Astley & Cummings (1984)) for acoustic transmission through ducts
with flexible walls. This study has shown that the radiation damping is negligible in
comparison to structural damping. Moreover, the investigation of the relation between
near field and far field has demonstrated the simplicity of the far field radiation pattern.
The internal reduction of sound in a duct with flexible walls and porous material has
been treated analytically by Astley (1990). A functional is derived based on the wave
equations in the airway, the porous material and the flexible structure. The dispersion
relationship obtained is solved to compute the propagating coupled wavenumber. The
imaginary part is converted to a loss factor.
Paper I focuses on two aspects of low frequency (plane wave) sound propagation in
rectangular ducts. The first one concerns the internal attenuation of sound waves inside
a lined duct with flexible walls. From a functional based on boundary conditions and
wave equations in the structure and fluid, the fundamental characteristics of sound in
flexible ducts are computed. The model used is based on Astley (1990).
The second aspect deals with the radiation of sound by the same duct. Using the
equations derived for the internal sound attenuation model, one can analytically estimate
the radiation. An analytical model, which regards the duct as a line of monopole
sources, is used. This procedure is applicable and gives good results as long as only the
fundamental plane mode is investigated. The model used in this case is based on
Cummings (1978, 1979).
These two theoretical models have been validated experimentally using different test
ducts. The two-microphone technique was used and radiation measurements were
performed in a reverberation room at the MWL laboratory. All measurements were
done in the plane wave range of the test ducts (0-2000 Hz).
One aim of this research is to implement the models in the software code SID (Sound
in ducts) developed at MWL to simulate sound propagation in duct networks (Nygrd
(2000).
2.1.1
Theoretical and measurement results of the internal transmission loss for a lined and
unlined corrugated duct are shown in Figure 2. In the lined case, only the bottom wall
was lined with polyurethane foam (thickness: 22 mm, length: 2m).
Unlined case
Lined case
Figure 2: Predicted and measured internal transmission loss for the unlined and lined
corrugated steel duct. Duct length: l=2m, cross-section area: 150 200 mm 2 , wall thickness:
1 mm.
The measurement results are in close agreement with data obtained from the theoretical
models. Both plots display four resonance peaks due to the presence of two pairs of
walls of different lengths (150 and 200 mm). At these frequencies maximum
transmission loss occurs. It can be seen that in the lined case (right plot), the trend is for
the transmission loss to increase as frequency rises. This is due to the presence of the
lining material which is more acoustically efficient for higher frequencies. No
noticeable deterioration between measurement and theory can be observed for
frequencies above the cut-off frequency of the corrugated duct (850 Hz) for which
higher-order mode perturbations would be expected. Inspection of the left plot (unlined
case) reveals a standing-wave like pattern. This can be explained by reflected waves
produced at the duct termination. Figure 3 shows the internal transmission loss for the
three tested duct
30
Rigid duct
Rectangular flexible duct
First duct (one flexible wall)
25
20
15
10
100
200
300
400
800
900
Figure 3: Comparison of measured internal transmission loss for the three lined ducts
investigated.
Except at the first bending wave resonances the lined flexible ducts behave as a rigid
duct.
Theoretical and measured results of the radiation transmission loss for the lined and
unlined corrugated duct are shown in Figure 4. As for the transmission loss analysis, the
bottom wall was lined with the same material (polyurethane foam).
Unlined case
Lined case
Figure 4: Predicted and measured radiation transmission loss for the unlined and lined corrugated
duct
Good agreement between theory and measurement is observed even for frequencies
above the corrugated duct cut-off frequency (850 Hz). This time, four dips are displayed
at the frequencies for which the internal transmission loss reaches a peak (resonance
frequencies of the flexible walls). The reason for this behaviour is that the acoustical
energy is mostly attenuated in the structure. No deviation occurs at resonance
frequencies although the radiation model based on monopole distribution was expected
to be violated. Comparison of the unlined and lined cases shows that the presence of the
lining does not greatly influence the radiation magnitude of the duct.
60
Experiment
Theory
Unlined
55
50
45
40
35
30
125
160
200
250
315
400
500
Frequency (Hz)
630
65
800
1000
Experiment
Theory
60
Lined
55
50
45
40
35
30
125
160
200
250
315
400
500
Frequency (Hz)
630
800
1000
2.2
The exhaust system of a car is typically composed of various elements such as the
silencer and termination (tailpipe). An efficient exhaust system implies the tuning of all
the elements respective to each other. The silencer has been exhaustively investigated.
The termination of the exhaust system has on the other hand received less attention.
A classical theoretical work on the sound radiation from on open end straight pipe with
no flow has been done by Lewine & Schwinger (1948). They obtain numerically the
reflection coefficient for the velocity potential and the amplitude of the diverging
spherical wave in the far field. The reciprocity relation is employed to relate the
radiation characteristics to the attenuation of the plane sound waves.
Another major work on the radiation of sound from an open ended straight pipe with
flow has been performed by Munt (1977, 1990). The mathematical model involves the
use of the Wiener-Hopf technique and a full Kutta condition (boundary conditions at the
lips stating that the flow velocity is fixed and finite). The model describes the air jet as a
semi-infinite cylinder pipe bounded by an unstable boundary layer. Cargill (1982)
derived analytical expressions for the far field radiation and pressure reflection
coefficient at the opening of a flow pipe. The agreement with Munts model is good.
The aeroacoustic behaviour of an open pipe has been treated experimentally with a
multimicrophone technique by Peters and al. (1993). The technique enables the
measurement of the acoustical attenuation due to visco-thermal effects along the wall
and the determination of the impedance, the pressure reflection coefficient and end
correction at the opening. The authors point out that the geometry of the opening does
not affect the low frequency reflection coefficient. The internal acoustic damping along
the walls of a straight pipe with an open end has also been treated by Dokumaci (1998),
Davies (1988) and Howe (1995). The attenuation in the core of the fluid is usually
neglected.
Flow diffusers and bends have been characterised by Dequand and al. (2002), van Lier
and al. (2001). A parallel has been drawn between the two geometries. It has been found
that the aeroacoustic behaviour is accurately predicted by a quasi-stationary model
10
when the flow separation occurs at sharp edges. Non-linear effects are more important
for smooth geometries and can not be predicted by an incompressible quasi-steady
theory. In the case of a smooth bend, it is theoretically preferable to replace the element
with an equivalent smooth straight pipe. The study of flow diffuser (van Lier and al.
(2001)) has shown that at low and moderate sound amplitude, the diffuser is a source of
sound at a specific Strouhal number. At high excitation level, the diffuser behaves as a
sound sink.
Paper II starts by describing the acoustics of a straight pipe with circular termination
which is used as a reference case. A second part treats modified tailpipes, i.e., a straight
pipe terminated with different configurations such as oblique cuts, diffusers and with
upstream bends
Straight pipe
11
(oblique cuts, diffusers and bends). Measured data are compared with theoretical
models based on Munt (1977, 1990). A low frequency experimental investigation has
shown that the diffuser elements deviate from the straight pipe case when flow is
present. It has been found that the magnitude of the pressure reflection coefficient for
the diffusers is larger than 1 and increase with the diffuser angle and flow Mach
number. The measured impedance data for flow diffusers are implemented in the
modelling of exhaust lines to test the potential of such elements in reducing sound
radiation.
Another aspect of interest is the determination of flow-induced noise emitted from the
same geometries; no other sources of sound being present. The main purpose being to
estimate the level of extra-noise generated by the elements compared to the reference
case (straight pipe). Based on this data, scaling laws could be obtained for the elements
after collapse of measured sound power levels as function of Strouhal number. The
velocity dependence found is of importance to understand the mechanism governing the
flow noise generation.
2.2.1
Straight pipe
Experimental results for internal damping in the boundary layer
The influence of flow on acoustic damping can be seen in Figure 7. The theoretical
damping presented in Figure 7 corresponds to the absolute value of the imaginary part of
the wavenumbers k+/- obtained from the cluster technique. The experimental results for
the no flow case in Figure 7(a) agree well with the classical Kirchoff (1868) formulation.
However, with flow (Figure 7(b)), the often quoted formulation by Davies (1988) does
not work and can significantly under predict the damping.
12
0.16
0.14
No flow case
Damping (1/m)
0.12
0.1
0.08
0.06
+ (Exp)
0.04
- (Exp)
0.02
0
50
Theory
500
1000
3000
3500
4000
(a)
+ (Exp)
Damping (1/m)
0.25
- (Exp)
+ (Theory)
0.2
- (Theory)
0.15
0.1
0.05
0
50
M=0.14
500
1500
2500
Frequency (Hz)
3500
4500
(b)
Figure 7: Acoustical attenuation rate in a straight pipe with and without flow.
13
Experimental data presented in Figure 8 are obtained using the cluster technique without
iteration (two-microphone technique, Bodn & bom (1986)) and a pure real
wavenumber k=/c (no damping) and with iteration. The latter corresponding to the
case for which the measured damping is included in the experimental determination of
the reflection coefficient R using the complete wavenumber k=/c-i, where is the
attenuation rate.
1.1
Munt
Measured damping (Exp)
No damping (Exp)
1
0.9
|R|
0.8
0.7
0.6
No flow case
0.5
0.4
0.1
0.4
0.6
0.8
1
ka
1.2
1.4
1.6
1.8
(a)
1.3
Munt
Measured damping (Exp)
No damping (Exp)
1.2
1.1
|R|
0.9
0.8
M= 0.18
0.7
0.6
4
St (ka/M)
(b)
Figure 8: Measured reflection coefficient and comparison with Munts model (1990).
14
Attenuation along the pipe does not greatly modify the reflection coefficient |R| when
no flow is present in the pipe (Figure 8(a)). As can be seen from Figure 8(b) the Munt
model agrees quite well with the measured data except for Strouhal numbers < 1.
End correction results
Munts model allows for the determination of the complex value of the reflection
coefficient R at the opening. In this section, the phase of R is converted into an endcorrection term. Both experimental and theoretical results are presented and shown in
Figure 9(a), (b).
Random excitation
Stepped-sine excitation
Munt
end correction
1
0.8
0.6
0.4
No flow case
0.2
0
0.23 0.4
0.6
0.8
1
1.2
ka
1.4
1.6
(a)
Figure 9 (a): End correction for an open end.
15
Experiment
Munt
end correction
1
0.8
0.6
0.4
0.2
M=0.1
0
1.5 3 4
8
10
St (ka/M)
12
14
16
(b)
Figure 9 (b): End correction for an open end.
For the no flow case, prediction of the end correction agrees well with the experimental
value. The low frequency limit obtained experimentally is very close to the value of
0.6133 derived by Levine & Schwinger (1948). An interesting and already noticed
feature is observed in the high Strouhal number region. In this portion of the plot
(St=ka/M >>1), one can see from comparison with the no flow case that the flow does
not change the value of the end correction. The theory tends to overestimate the end
correction at low Strouhal numbers (St< 5). It can also be seen that the measurements
clearly support the drop in the end correction for small St-numbers predicted by Munts
model. For small St-numbers the flow field plays an active part and coupling between
the acoustic field and the flow field is of importance. This coupling is imposed by Munt
theoretically through the full Kutta condition. This condition governs the transfer of
energy between the acoustic field and the flow field.
Sound absorption by vorticity shedding
The presence of mean flow in the pipe influences the rate of sound radiated to the far
field. The results presented in Figure 5 show that acoustic losses due to vorticity is
16
dominantly a low frequency process. The upper frequency limitation of the theory
developed by Bechert (1980) is indicated by a vertical dashed line in Figure 10.
20
10*log(Wrad/Wtrans)
10
0
-10
-20
-30
M=0.1
-40
-50
-60
Measurement
BECHERT
Theory validity
-70
0
1000
2000
3000
Frequency (Hz)
4000
5000
20
10log(Wrad/Wtrans)
10
0
-10
-20
-30
M=0.15
-40
-50
-60
-70
0
Measurement
BECHERT
Theory validity
500
1500
2416
3500
Frequency (Hz)
4500
Figure 10: Theoretical (Bechert (1980)) and measured sound absorption by vorticity.
17
Tailpipes
Reflection coefficient
The experimental and theoretical results presented in this section have been obtained at
the inlet of the elements.
Oblique cuts
Figure 11 presents the oblique cut dependence for the magnitude |R| of the reflection
coefficient for the case of no flow. Theoretical models for one oblique cut case (=45,
M=0) are validated against measurement in Figure 12.
1.5
Straight pipe
45
60
30
1
|R|
0.5
No flow case
M=0
0.5
1
Helmholtz ka
1.5
Figure 11: Reflection coefficient for oblique cuts in terms of Helmholtz number for no flow
case.
18
1.5
=30
|R|
0.5
M=0
0
150
1000
2000
Frequency (Hz)
3000
4000
Figure 12: Models for reflection coefficient for an oblique cut (45) and no flow.
Inspection of Figure 11 reveals the major feature of these geometries. The reflection
coefficient magnitude drops and the transmission increases. This can have a positive
effect since one problem with normal tailpipes is resonances, which amplify the radiated
sound at certain frequencies. For a fixed frequency the transmission increases with the
angle and for a fixed angle the effect increases with the frequency. Figure 12 shows that
increasing the cross-section area, i.e, using an equivalent radius at the opening is not
sufficient to predict the reflection coefficient. In other words, the obliquity has also an
effect on the acoustical propagation. To take this geometrical effect into account, a
model based on parallel coupled pipes with different lengths was developed.
19
Diffusers
Figure 13 and Figure 14 illustrate the diffuser case for two angles 5 and 7.5 for the no
flow case and a case with flow.
1.4
Experiment
SID
1.2
1
0.8
|R|
=5
0.6
No flow case
0.4
0.2
Figure 13: Experimental and theoretical reflection coefficient for diffuser (5) for the no flow
case.
1.4
Experiment
SID
1.2
1
=7.5
|R|
0.8
0.6
0.4
M=0.09
0.2
0
500
Figure 14: Experimental and theoretical reflection coefficient for diffuser (7.5) and M= 0.09.
20
The simple model used to predict the reflection coefficient for the no flow case agrees
well with the experimental data. The model is based on the conical element in the SID
code2 terminated by an impedance obtained from Munts theory (1977, 1990).
Modelling of the standing wave pattern (undulation) is achieved theoretically. The
discrepancy between the calculated and measured values for lower frequency when flow
is present in the rig (Figure 14) is suspected to be due to the flow separation that occurs
for this angle. Dequand and al. (2002) showed for diffusers of this type the flow
separation point is not fixed and moved along the inner walls of the elements. Munts
model used at the outlet cross-section is based on a sharp-edged condition with a thin
shear boundary layer. Due to the flow separation the boundary layer at the opening is
substantially thicker violating the simple flow separation model used by Munt.
Bends
The reflection coefficient results for tailpipes with an upstream bend confirms the
findings of Dequand and al. (2002). The bends can be replaced theoretically by a
straight pipe.
Flow noise
Oblique cuts
Figure 15 draws an experimental comparison for the flow induced sound power level
for the oblique cut. Only the last case corresponding to the most pronounced angle (60)
is shown since it is obviously the loudest. It is compared to the measurement with no
element (straight pipe reference case). No significant difference in terms of the sound
21
flow production can be observed for the oblique cuts. It behaves like a straight-pipe; no
additional noise is generated due to the obliquity.
80
Straight pipe
70
M=0.19
Oblique cut 60
Lw (dB)
60
M=0.1
50
40
30
M=0.06
500
Figure 15: Measured sound flow power level for oblique cut (60) and straight pipe.
Dimensionless sound power spectra for diffusers and bends are presented in Figure 16.
The dimensionless spectra Ls = 10 log10 (F (St )) , where St is the Strouhal number based
on the pipe diameter for the bends and outlet diameter for the diffusers is based on the
equation
W = 0U 3 D 2 M CL2 F ( St )
(2-1)
where W is the flow induced noise sound power, U is the flow speed, D is the pipe
diameter. The exponent depends on the source type and the dimensionality of the
sound field.
The velocity dependence for the bends is U 4 and U 6 for the diffusers (1/3-octave
band). As expected bend 1 corresponding to the greatest curvature (R/D=1.6) generates
higher sound power level.
22
-30
Diffuser 7.5
-35
-40
Ls (dB)
Diffuser 5
-45
Diffuser 2.5
-50
-55
-60
-65
Straight pipe
-0.3
-0.2
-0.1
0
0.1
log(St)
0.2
0.3
0.4
-40
Mitre bend (Nygrd)
-45
Ls (dB)
-50
-55
90 bend R/D=1.6
-60
-65
90 bend R/D=2.5
-70
45 bend L=5D
-75
-0.4
-0.2
0.2
0.4
0.6
0.8
log(St)
Figure 16: Complete sets of dimensionless spectra based on equation (2-1) obtained from our investigation
and Nygrd (2000).
23
Oblique cuts
Deviations from the straight pipe at 2.5, 5, 7.5: Flow noise sound power
Bends
level at opening U 6
45 bend(L=0): Equivalent to a
straight pipe
45 bend(L=5D), 90bends (R/D=1.6,
The measured load impedances for flow diffusers have been implemented in the
calculation of the insertion loss (IL) of exhaust lines. Figure 17 demonstrates the
potentiality of flow diffusers to increase the losses at the tailpipe resonances.
7
6
5
M=0.06
M=0.1
M=0.2
4
IL (dB)
3
2
1
0
-1
-2
-3
20
60
80
100
Frequency (Hz)
120
140
Figure 17: Increase in IL for the diffuser 7.5 at various flow Mach numbers M.
24
2.3
Paper III. Linear and non linear acoustic regimes of perforated plates
Sound absorbers are traditionally made of porous materials of fibrous types. The
drawbacks of these types of materials are numerous: they are brittle therefore
hazardous, bulky and inefficient at low frequencies. A new trend in the automotive
industry is to use thin micro-perforated panels to enhance the noise reduction. These
panels are required to be made with perforations of sub-millimetre size in order to
obtain high acoustic attenuation without the drawbacks of the traditional porous
materials. An analytical treatment has been suggested by Maa (1998) to model the ratio
of the acoustic pressure drop through a panel made of micro-circular holes to the
acoustic particle velocity over the panel. This ratio is called the acoustic (flow)
impedance and is used to characterise the material. Based on Maas formulation, a
similar analytical expression for panels made of rectangular apertures is derived in
paper III.
Measurement of the impedance of single rectangular and circular apertures at low and
high excitation amplitudes has been performed by Sivian (1935). An analytical model
consisting of the various internal and external linear acoustics terms of an aperture is
also detailed. However, only non linear resistive results are presented. Sivian (1935)
neglects the role of the reactance on the overall acoustics of the opening.
Ingard (1967, 1974) has investigated the acoustics of apertures. Numerous effects such
as interaction effects between neighbouring apertures, non linear effects, the effects of
the geometry of the plane (baffled, unbaffled, circular, rectangular) in which the
aperture is located are treated by Ingard.
Randeberg (2000, 2002) has derived a model for a slitted plate with multiple layers. The
analytical model is coupled to a finite element scheme to predict the absorption
coefficient of the complete assembly. The modelling is checked with a slitted sample
mounted in an impedance tube. In paper III various experiments are carried out on
perforated samples with single or multiple apertures of square, circular and rectangular
geometries mounted in a rig to measure the linear and non-linear impedance. Empirical
observations and experimental comparisons between the different plates investigated are
made. A technique is suggested to measure the perforation area ratio (porosity) of
these samples.
25
2.3.1
Figure 19 shows the flow impedance of single apertures of square and rectangular
geometries. It can be seen that the more elongated the aperture, the lower is the
reactance due to that the end correction term is reduced.
Figure 20 demonstrates the differences between square, rectangular and circular
geometries with the same cross-sectional area. It can be seen that the square and circular
apertures have the same acoustics behaviour. An elongation of the aperture towards a
rectangular geometry results in the reduction of the end correction as observed for
Figure 19.
26
Normalised resistance
based on the particle velocity inside the aperture
0.045
0.04
0.035
0.03
0.025
0.02
100 Hz
0.015
0.01
0.005
0
0
4
6
8
10
Peak particle velocity inside the aperture (m/s)
Normalised reactance
based on the particle velocity inside the aperture
0.018
12
0.016
0.014
0.012
100 Hz
0.01
0.008
0.006
0.004
0.002
0
0
4
6
8
Peak particle velocity inside the aperture (m/s)
10
12
Figure 19: Normalised flow impedance S based on the interior particle velocity V for plates with
single square and rectangular apertures.
27
0.045
Normalised resistance
based on the particle velocity inside the aperture
0.04
w=11.4 mm, h=11.4 mm
0.035
d=12.9 mm
0.03
100 Hz
0.025
0.02
0.015
100 Hz
300 Hz
0.01
300 Hz
0.005
300 Hz
0
0
4
6
8
10
Peak particle velocity inside the aperture (m/s)
Normalised reactance
based on the particle velocity inside the aperture
0.05
14
300 Hz
0.045
12
0.04
d=12.9
mm
0.035
300 Hz
0.03
0.025
0.02
0.015
100 Hz
100 Hz
0.01
100 Hz
0.005
0
4
6
8
Peak particle velocity inside the aperture (m/s)
10
12
Figure 20: Normalised impedance S based on the interior particle velocity V for plates with
single square, rectangular and hole orifices ( =0.0265).
28
Figure 21 shows the impedance for samples composed of multiple apertures with the
same geometry (width w: 5 mm, height h: 1 mm). Interaction effects are observed as the
number of slits increases. A comparison between the linear slit model (straight line) and
linear and non-linear measurement carried out on a slitted plate is given in Figure 22 and
shows good agreement.
Normalised resistance
based on the particle velocity inside the aperture
0.02
0.018
0.016
100 Hz
0.014
0.012
0.01
2 slits
4 slits
16 slits
0.008
0.006
0.004
0.002
0
0
3
4
5
6
7
8
Peak particle velocity inside the aperture (m/s)
10
-3
Normalised reactance
based on the particle velocity inside the aperture
5.5
x 10
2 slits
4 slits
16 slits
100 Hz
4.5
3.5
2.5
2
0
2
3
4
Peak particle velocity inside the aperture (m/s)
Figure 21: Single aperture normalised flow impedance S for plates with multiple rectangular
apertures (w=5 mm, h=1 mm).
29
1.5
Normalised resistance
100 Hz
0.5
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Peak particle velocity on the sample inlet surface (m/s)
0.18
0.2
0.18
0.2
1
0.9
100 Hz
Normalised reactance
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Peak particle velocity on the sample inlet surface (m/s)
Figure 22: Normalised flow impedance S for a plate with sixteen rectangular apertures (w=5 mm,
h=1 mm, =0.016).
30
Main contributions
Paper I: A new two-port model for rectangular flexible ducts has been developed.
Paper II: First systematic investigation of tailpipe aeroacoustics.
Paper III: A new analytical model for the acoustic impedance of slitted plates has been
derived.
The papers also contain a number of results and conclusion of practical interest.
Paper I has demonstrated that except at the first wall bending wave resonance the effect
of flexible walls can be neglected in practice. It has been shown that the lining position
is not of importance in the low frequency (plane wave) range..
Paper II has shown that an outlet diffuser can significantly modify the low frequency
reflection coefficient and reduce the effect of low frequency tailpipe resonances. An
oblique cut tailpipe opening can reduce the reflection coefficient at high frequencies and
reduce the effect of high frequency tailpipe resonances. Other results of the study are
summarised in Table 1.
Paper III has presented an experimental technique to obtain the porosity of perforated
plates.
31
4 Future research
Paper I. Low frequency sound propagation in rectangular ducts
To improve the presented two-port model, in particular around the cut-on frequencies
for the structural bending waves, a complete 1-D wave model could be derived. Such a
model should include not only the acoustic wave types but also the structural bending
waves and could be build on the works presented by Martin (1991) and Martin and al.
(2004). Another topic of interest is to investigate the effect of flow and the existence of
instability waves. Such waves could exist close to the cut-on frequency of the bending
waves where the wall impedance is close to zero.
Paper II. Aeroacoustic behaviour of tailpipes
The model of Munt (1977, 1990) has been coded to predict the acoustics at the opening.
However, in its full application, sound absorption in the near field and the sound
radiation can also be calculated. This could be used for a more complete validation of
the dissipation by vorticity at the opening and to investigate the far-field directivity. It
has been shown that flow diffusers can reduce the sound radiation at low frequencies by
reducing the effect of tailpipe resonances. However, a diffuser will also increase the
flow noise at low and mid frequencies. A more detailed experimental study of these two
effects would reveal the true potential of this type of acoustic elements. Non-linear
effects could be also investigated since high noise amplitude propagates in exhaust
systems.
Paper III. Linear and non linear acoustic regimes of perforated plates
Here the work should focus on developing models for non-linear impedance effects for
rectangular (slit) apertures. Also the effect of grazing flow on the impedance of microperforated plates is important for vehicles applications. A large amount of works exist
for perforated pipes and aircraft liners (Elnady (2004)). These works present semiempirical models, but the data are based on plates with larger holes and higher
perforation ratios than what is typical for micro-perforated plates.
32
5 References
1. bom M., 1989, Studies of sound propagation in ducts and mufflers, Doctoral Thesis,
Department of Aeronautical and Vehicle Engineering, Royal Institute of Technology,
Stockholm, Sweden.
2. bom M., 1999, Determination of porous material data via two-port measurements, InterNoise 99, Florida, USA, 589-594.
3. bom M., Bodn H., 1988, Error analysis of two-microphone measurements in ducts with
flow, Journal of the Acoustical Society of America, 83, 2429-2438.
4. Allam S., 2004, Acoustic modelling and testing of advanced exhaust system components for
automotive engines, Doctoral Thesis, Department of Aeronautical and Vehicle Engineering,
Royal Institute of Technology, Stockholm, Sweden.
5. Allard J.F., 1993, Propagation of sound in porous media, modelling sound absorbing
materials, Elservier, London.
6. Almgren M, 1982, Prediction of sound pressure level outside a closed ventilation duct a
literature survey, Report F82-03, Chalmers University of Technology, Department of Building
Acoustics.
7. Astley, R. J., 1990, Acoustical modes in lined ducts with flexible walls. A variational
approach, Proceedings of Inter-Noise 90, 575-578.
8. Astley R. J., 1991, A finite element scheme for acoustic propagation in flexible-walled ducts
with bulk-reacting liners, and comparison with experiment, Journal of Sound and Vibration,
150(1), 119-138.
9. Astley R.J., Cummings A., Sormaz N., 1991, A finite element scheme for acoustical
propagation in flexible walled ducts with bulk reacting liners, and comparison with experiment,
Journal of Sound and Vibration, 150, 119-138.
33
10. Astley R. J., Cummings A., 1984, A finite element scheme for acoustic transmission through
the walls of rectangular ducts: comparison with experiment, Journal of Sound and Vibration,
92(3), 387-409.
11. Astley R.J., Cummings A., 1987, A finite element scheme for attenuation in ducts lined with
porous material: comparison with experiment, Journal of Sound and Vibration, 116(2), 239263.
12. Baxa D. E., 1989, Noise control in combustion engines, Krieger Pub Co.
13. Bechert D. W., 1980, Sound absorption caused by vorticity shedding, demonstrated with a
jet flow, Journal of Sound and Vibration, 70, 389-405.
14. Bierkens J., 2002, Calculation of the reflection coefficient of a jet pipe with subsonic jet
flow, Master Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of
Technology, Stockholm, Sweden.
15. Blevins R. D., 1984, Applied fluid dynamics handbook, New York: Van Nostrand Reinhold.
16. Bodn H., 1989, Characterization of fluid machines as sources of fluid borne noise,
Doctoral Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of
Technology, Stockholm, Sweden.
17. Bodn H., bom M., 1986, Influence of errors on the two-microphone method for
measuring acoustic properties in ducts, Journal of the Acoustical Society of America, 79(2),
541-549.
18. Bodn H., Carlsson U., Glav R., Wallin H. P., bom M., 2001, Ljud och vibration,
Norstedts Tryckeri AB, Stockholm.
19. Boij S., 2003, Acoustic scattering in ducts and influence of flow coupling, Doctoral Thesis,
Department of Aeronautical and Vehicle Engineering, Royal Institute of Technology,
Stockholm, Sweden.
34
20. Brown G. L., Rennison D. C., 1974, Sound radiation from pipes excited by plane acoustic
waves, Noise, Shock & Vibration Conference, Monash University, Melbourne.
21. Callow G.D., Peat K.S., 1988, Insertion loss of engine intake and exhaust silencers, IMechE
C19/88, 39-46.
22. Cargill A. M., 1982, Low frequency acoustic radiation from a jet pipe- a second order
theory, Journal of Sound and Vibration, 83, 339-354.
23. Cabelli A., 1985, The propagation of sound in a square duct with a non-rigid side wall,
Journal of Sound and Vibration, 103(3), 379-394.
24. Cabelli A., 1985, Application of the time dependent finite difference theory to the study of
sound and vibration interactions in ducts, Journal of Sound and Vibration, 103, 13-23.
25. Chung J.Y., Blaser D.A., 1980, Transfer function method of measuring in-duct acoustic
properties I. Theory, Journal of the Acoustical Society of America, 68, 907-913.
26. Crandall I. B., 1927, Theory of vibrating systems and sound, Appendix A: Resistance
coefficients for cylindrical conduits, Macmillan and co, limited.
27. Cremer L., Mller H. A., 1976/1982, Principles and applications of room acoustics, vol. II,
Applied Science Publisher, London.
28. Cummings, A., 1976, Sound attenuation in ducts lined on two opposite walls with porous
material, with some applications to splitters, Journal of Sound and Vibration, 49(1), 9-35.
29. Cummings, A., 1978, Low frequency acoustic transmission through the walls of rectangular
ducts, Journal of Sound and Vibration, 61(3), 327-345.
30. Cummings, A., 1979, Low frequency sound transmission through the walls of rectangular
ducts: further comments, Journal of Sound and Vibration, 63(3), 463-465.
31. Cummings, A., 1979, The effects of external lagging on low frequency sound transmission
through the walls of rectangular ducts, Journal of Sound and Vibration, 67(2), 187-201.
35
32. Cummings, A., 1980, Low frequency acoustic radiation from duct walls, Journal of Sound
and Vibration, 71(2), 201-226.
33. Cummings, A., 1981, Design charts for low frequency acoustic transmission through the
walls of rectangular ducts, Journal of Sound and Vibration, 78, 269-289.
34. Cummings, A., 1981, Stiffness control of low frequency acoustic transmission through the
walls of rectangular ducts. Journal of Sound and Vibration, 74, 351-380.
35. Cummings, A. 1983, Approximate asymptotic solutions for acoustic transmission through
the walls of rectangular ducts, Journal of Sound and Vibration, 90(2), 211-227.
36. Cummings, A., 1983, Higher order mode acoustic transmission through the walls of
rectangular ducts, Journal of Sound and Vibration, 90(2), 193-209.
37. Cummings, A. 1994, The attenuation of sound in unlined ducts with flexible walls, Journal
of Sound and Vibration, 174(4),
38. Cummings, A., 2001, Sound transmission through duct walls, Journal of Sound and
Vibration, 239(4), 731-765.
39. Cummings A., Astley R.J., 1995, The effects of flanking transmission on sound attenuation
in lined ducts, Journal of Sound and Vibration, 179, 617-646.
40. Davies P.O.A.L., 1988, Practical flow duct acoustics, Journal of Sound and Vibration,
124(1), 91-115.
41. Davies P.O.A.L., Coelho J. L. B., Bhattacharya M., 1980, Reflection coefficients for an
unflanged pipe with flow, Journal of Sound and Vibration, 72, 543-546.
42. Dequand S., van Lier L., Hirschberg A., Huijnen J., 2002, Aeroacoustics response of
diffusers and bends: comparison of experiments with quasi-steady incompressible flow models,
Journal of Fluids and Structures, 16, 957-969.
43. Dokumaci E., !998, On transmission of sound in circular and rectangular narrow pipes with
superimposed mean flow, Journal of Sound and Vibration, 210, 375-389.
36
44. Ducret F., 2005, Aeroacoustics behaviour of a straight pipe, Twelfth International Congress
on Sound and Vibration, Lisbon, Portugal.
45. Ducret F., bom M., 2004, Aeroacoustics behaviour of different exhaust pipe geometries,
Eleventh International Congress on Sound and Vibration, St. Petersburg, Russia.
46. Ducret F, bom M., 2003, Low frequency sound propagation in rectangular ducts, Tenth
International Congress on Sound and Vibration, Stockholm, Sweden.
47. Ducret F., Jacobsen F., bom M., 2005, Development of micro-perforated acoustic elements
for vehicle applications, Twelfth International Congress on Sound and Vibration, Lisbon,
Portugal.
48. Elnady T., 2002, On the use of liners to reduce aircraft jet engine noise, Licenciate Thesis,
Royal Institute of Technology, Department of Aeronautical and Vehicle engineering, TRITAFKT 2002:26.
49. Elnady T., 2004, Modelling and characterization of perforates in lined ducts and mufflers,
Doctoral Thesis, KTH Aeronautical and Vehicle Engineering.
50. Fenech B, 2004, Damping the acoustic modes in a closed cavity using a micro-perforated
plate, MSc Thesis, Acoustic Technology, rsted, DTU, Denmark.
51. Gijrath H., Nygrd S., bom M., 2001, Modelling of flow generated sound in ducts, Eight
International Congress on Sound and Vibration, Hong-Kong, China.
52. Glav R., 1994, On acoustic modelling of silencers, Doctoral thesis, Royal Institute of
Technology, Department of Aeronautical and Vehicle engineering, TRITA-FKT Report 9435.
53. Goldstein M., 1976, Aeroacoustics, McGraw-Hill, New York.
54. Gomperts M.C., 1964, The sound insulation of circular and slit-shaped apertures,
Acustica, 14(1), 1-16.
37
55. Gomperts M.C., 1965, The influence of viscosity on sound transmission through small
circular apertures in walls of finite thickness, Acustica, 15(4).
56. Gomperts M.C, Kihlman T., 1967, The sound transmission loss of circular and slit-shaped
apertures in walls, Acustica, 18, 144-150.
57. Hirschberg A., Bruggeman J. C., Wijnands A. P. J., Morgenstern M., 1988, The whistler
nozzle and horn as aero-acoustic sound sources in pipe systems, Proc. Inst. Acoust., 10, 701708.
58. Howe M. S., 1980, The dissipation of sound at an edge, Journal of Sound and Vibration, 70,
407-411.
59. Howe M. S., 1995, The damping of sound by turbulent shear layers, Journal of the
Acoustical Society of America, 98, 1723-1730.
60. Huijnen J.H., 1998, Aero-acoustics of a bend: quasi-stationary models, Technische
universiteit eindhoven, Report R1476A.
61. Ingard U., 1953, On the Theory and Design of Acoustic Resonators, Journal of the
Acoustical Society of America, 25(6), 1037-1061.
62. Ingard U., Ising H., 1967, Acoustic nonlinearity of an orifice, Journal of the Acoustical
society of America, 42(1), 6-17.
63. Ingard U., Singhal V. K., 1974, Sound attenuation in turbulent pipe flow, Journal of the
Acoustical Society of America, 55, 535-538.
64. Ingerslev F., Kjerbye Nielsen A., 1944, On the transmission of sound through small
apertures and narrow slits, Publication Nr.1 from The Acoustical Laboratory of the Academy
of Technical Sciences, Copenhagen.
38
65. int panhuis P., 2003, Calculations of the sound pressure reflection coefficient and the
acoustic end correction of a semi-infinite pipe issuing a subsonic cold or hot jet with co-flow,
Master Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of
Technology, Stockholm, Sweden.
66. ISO 3741, 1994, Acoustics: Determination of sound power levels of noise sources using
sound pressure, precision methods for reverberation rooms.
67. ISO 3747, 1998, Acoustics: Determining the sound power level of noise sources from sound
pressure measurement.
68. Jacobsen F., 2005, Propagation of sound waves in ducts, Note 31260, Acoustic Technology,
rsted-DTU, Denmark.
69. Kinsler L. E., Frey A. R., Coppens A. B., Sanders J. V., 1982, Fundamentals of acoustics,
John Wiley & Sons, Inc
70. Kirby R., Cummings A., 1998, Structural/acoustic interaction in air-conditioning ducts in
the presence of mean flow, Proceedings of ISMA23, Leuven, Belgium, 677-684.
71. Kirchoff G., 1868, Uber den einfluss der wrmteleitung in einem gase auf die
schallbewegung, Pogg. Ann., 134(6), 177-193.
72. Kristiansen U.R., Vigran T.E., 1994, On the design of resonant absorbers using a slotted
plate, Applied Acoustics, 43, 39-48.
73. Kuhn G.F., Morfey C.L., 1976, Noise due to fully developed turbulent flow exhausting from
straight and bend pipes, Journal of Sound and Vibration, 44, 27-35.
74. Lamb H., 1910, The dynamical theory of sound, London.
75. Levine, H., Schwinger H., 1948, On the radiation of sound from an unflanged circular pipe,
Physical Review, 73, 383-406.
39
76. Maa D-Y., 1983, Direct and accurate impendence measurement of microperforated panel,
Inter Noise, Edinburg, UK.
77. Maa D-Y., 1987, Microperforated-panel wideband absorbers, Noise Control Engineering
Journal, 29(3).
78. Maa D-Y., 1994, Microperforated panel at high sound intensity, Inter Noise, Yokohama
Japan, August 29-31.
79. Maa D-Y., 1998, Potential of microperforated panel absorber, Journal of the Acoustical
Society of America, 105(5), 2861-2866.
80. Martin V., 1991, Perturbation of fluid-guided waves introduced by bending plates, Journal
of Sound and Vibration, 144, 331-353.
81. Martin V., Cummings A., Gronier C., 2004, Discrimination of coupled structural/acoustic
duct modes by active control: principles and experiments results, Journal of Sound and
Vibration, 274, 583-603.
82. Mechel F.P., 1994, Helmholtz resonators with slotted neck plates, Acustica, 80, 321-331.
83. Melling T.H., 1973, The acoustic impedance of perforates at medium and high sound
pressure levels, Journal of Sound and Vibration, 29(1), 1-65.
84. Morfey C.L, 1969, Acoustic properties of openings at low frequencies, Journal of Sound and
Vibration, 9(3), 357-366.
85. Morfey C.L., 1971, Sound transmission and generation in ducts with flow, Journal of Sound
and Vibration, 14(1), 37-55.
86. Morse, P.M., Ingard K.U., 1968, Theoretical acoustics, McGrawHill, St.Louis.
87. Munt R.M., 1977, The interaction of sound with a subsonic jet issuing from a semi-infinite
cylindrical pipe, Journal of Fluid Mechanics, 83, 609-61040.
40
88. Munt R. M., 1990, Acoustic transmission properties of a jet pipe with subsonic jet flow: I.
the cold jet reflection coefficient, Journal of Sound and Vibration, 142, 413-436.
89. Nelson P.A., Morfey C.L., 1981, Aerodynamic sound production in low speed flow ducts,
Journal of Sound and Vibration, 79, 263-289.
90. Nygrd S, 2000, Modelling of low frequency sound in duct networks, Licenciate Thesis,
Royal Institute of Technology, Department of Aeronautical and Vehicle engineering, TRITAFKT 2000:57.
91. Oldham D. J., Ukpoho A. U., 1990, A pressure-based technique for predicting regenerated
noise levels in ventilation systems, Journal of Sound and Vibration, 140, 259-272.
92. Peters M.C.A.M., Hirschberg A., Reijnen A.J., Wijnands A.P.J, 1993, Damping and
reflection coefficient measurements for an open pipe at low Mach and Low Helmholtz numbers,
Journal of Fluid Mechanics 256, 499-534.
93. Pierce A.D., 1989, Acoustics: an introduction to its physical principles and applications,
Acoustical Society of America, Mc Graw Hill Inc., New York.
94. Regaud P. L., 2001, Modelling of compact resonators, Licenciate Thesis, Royal Institute of
Technology, Department of Aeronautical and Vehicle engineering, TRITA-FKT 2001:31.
95. Lord Rayleigh, 1878, The theory of sound, Macmillan and Co.
96. Randeberg R.T., 2000, A Helmholtz resonator with a lateral elongated orifice, Acta
Acustica/Acustica, 86, 77-82.
97. Randeberg R. T., 2002, Adjustable slitted panel absorber, Acta Acustica/Acustica, 88, 507512.
98. Rienstra S.W., Hirschberg A., 2004, An introduction to acoustics, Report IWDE 01-03,
Eindhoven University of Technology, can be downloaded at:
http://www.win.tue.nl/%7Esjoerdr/papers/boek.pdf
41
99.
Ronneberger
D.,
1975,
Genaue
messung
der
schalldmpfung
und
der
VDI3733,1996,
Noise
at
pipes,VDI
manual
noise
reduction
guidelines,
42
Table of Contents
1. Introduction .................................................................................................................. 1
2. Theory........................................................................................................................... 5
2.1
Governing equations, functional and dispersion relationship .......................... 5
2.1.1
Air-borne waves ....................................................................................... 6
2.1.2
Waves in the liner ..................................................................................... 6
2.1.3
Structure-borne waves .............................................................................. 7
2.1.4
Boundary conditions................................................................................. 8
2.1.5
Derivation of the functional...................................................................... 9
2.1.6
Analytical formulation.............................................................................. 9
2.2
Flexible duct transfer matrix formulation....................................................... 11
2.2.1
Theoretical duct impedance.................................................................... 12
2.3
The line source method for acoustic radiation/breakout from duct walls ...... 13
3. Methods and materials................................................................................................ 17
3.1
Rig description................................................................................................ 17
3.1.1
Types of ducts......................................................................................... 17
3.1.2
Electronic equipment.............................................................................. 22
3.2
Internal experimental measurement techniques ............................................. 22
3.2.1
Experimental duct transfer matrix .......................................................... 22
3.2.2
Incident power and duct internal sound power....................................... 23
3.3
Duct radiated sound power (external measurement) ...................................... 25
4. Results and discussion ................................................................................................ 27
4.1
In-duct investigation ....................................................................................... 28
4.1.1
Duct with one flexible wall (internal propagation) ................................ 28
4.1.2
Corrugated duct (Internal propagation) .................................................. 30
4.1.3
Internal sound propagation for the rigid walled duct ............................. 32
4.2
Sound radiation investigation ......................................................................... 33
4.3
Parameter investigations................................................................................. 35
4.3.1
Effect of liner positioning....................................................................... 35
4.3.2
Effect of structural conditions ................................................................ 36
4.3.3
Effect of the reflection in the airway ...................................................... 37
5. Conclusions ................................................................................................................ 39
6. References .................................................................................................................. 41
Appendix A A note on measurement of liner properties ............................................ 45
1. Introduction
HVAC (heating, ventilation and air-conditioning) systems are commonly found in
buildings and vehicles. They are used to transport airflows to ensure comfortable
working or living conditions. The ducts used in HVAC systems have either a circular or
a rectangular cross section. For the case of rectangular ducts, wall vibrations are easily
excited and radiate sound. The walls are sometimes lined with porous material that
provides efficient acoustical attenuation at medium and high frequencies. Noise
produced by the units (fans, compressor) or picked up along the system from other
sources such as speech (breakin effects), propagates along the entire system and is
radiated (breakout effects) into other locations. These ventilation systems also carry
airflows with speeds up to 10 m/s corresponding to a Mach number M less than 0.05.
This implies that mean flow effects (convection) can be neglected in the analysis.
For high frequencies, well above the plane wave range, power based analysis methods
as described in standard procedures, e.g., ASHRAE (2003). However, such methods do
not apply to the low frequency plane wave region where standing wave effects can be
important. For the plane wave range, the most efficient is to model the duct system as a
network of acoustical 2-ports (Munjal (1987), Davies (1988)). Each element in a
network is characterised by its acoustic 2-port. The building block technique consists of
combining these elements to form complex networks of elements. The overall
performance of the entire system can then be calculated.
The aim of this study is to investigate low frequency propagation in rectangular ducts
and transmission through the walls and derive analytical models for implementation in
2-port codes, e.g., the SID code developed at KTH (Nygrd (2000)).
Various analytical and numerical methods have been developed (Astley (1990, 1991),
Astley & Cummings (1984, 1987), Cabelli (1985a), Cummings (1978, 1979b) to predict
the propagation of sound in ducts with flexible walls. A time-dependent finite difference
method (Cabelli (1985b) has been used. The variational technique has been employed by
Cummings & Astley (1995). A finite element technique has been used by Astley &
Cummings (1987), Astley and al. (1991), Martin (1991) and Kirby & Cummings
(1998).
Astley (1991) derives a dispersion relationship for the case of infinite-length lined
elastic ducts. The formulation starts with deriving the governing equations in the duct
cross-section in the different media (structure, airway, liner). The boundary conditions
of the geometry gives a functional which can form the basis for various theoretical
techniques and is defined for the low frequency range of interest corresponding to the
frequency range below the cut-off frequency of the duct. A uniformly distributed
pressure profile (plane wave) is assumed over the duct cross section.
Authors such as Cabelli (1985a) uses a Fourier series to define the pressure distribution
on the duct wall due to the presence of higher-order modes. A comparison of both
models of internal sound field (non-uniform in Cabelli and uniform in Cummings) has
been made by Cummings (2001). It turns out that up to the cut-on frequency of the
equivalent rigid duct, no deviation is observed between these two ways of expressing
the internal sound field acting on the flexible walls. Thus, it is not greatly in error to
consider only uniform pressure over the duct cross-section.
With this simplifying assumption, the minimisation of the functional is straightforward
and leads directly to a dispersion relationship. Implementation of an iterative technique
such as the Newton-Raphson scheme for solving the roots of this dispersive equation is
done to obtain the propagating coupled wavenumber along the duct axis. The novelty in
this paper is the derivation of the transfer matrix after theoretical derivation of the duct
wavenumber and impedance. The internal transmission loss is obtained from the
elements of this transfer matrix.
The analytical approach taken in this paper to obtain the radiated sound power level and
transmission loss has been derived by Cummings (1979b, 1980) based on the line source
model from Rennisson & Brown (1974). It regards the vibrating duct as a distribution of
monopole sources of infinite length. A radiation efficiency is introduced to compensate
for the finite length of the duct.
A finite element based analysis of the sound radiation from finite length rectangular
ducts has been performed by Astley & Cummings (1984).
The experimental apparatus used in this thesis consists of a broadband noise excitation
generated from a loudspeaker located on each side of the tested rectangular ducts. Using
an experimental technique involving measurement of the acoustical upstream and
downstream1 conditions, one can experimentally obtain the transfer matrix (bom
(1999)). No measurement is conducted inside the rectangular duct. In contrast,
Cummings (1978) used a traverse probe to scan the sound field inside the duct as done
in an impedance tube apparatus.
Radiation effects (breakout) are measured from a rotating microphone located in a
reverberation chamber. Liner properties required in the full prediction is measured
separately in a special rigid walled duct. All the measurements are done with no flow.
Noise retransmitted to the inside after reflection from the surrounding (breakin-breakout
bypass effect) is not treated.
The paper structure can be broken into two main parts: the first one accounts for the
internal propagation of the plane wave and the second part treats the radiation. Both
aspects are detailed theoretically with derivation of simple analytical models and
confronted with results from different experiments. Various experimental techniques are
described to characterise the sound fields in the different regions of the rig (circular pipe
sections, internal flexible duct and external region). Experimental investigations of the
termination effects in the airway and in the structure are also presented.
It should be noted that the upstream and downstream sections refer to the circular pipes placed on both
sides of the test ducts (see Figure 9). No airflow has been used in the paper.
2. Theory
In this section, mathematical derivations are presented for two theoretical models
investigated in this paper. The first one concerns the propagation of sound waves inside
a flexible walled duct and is based on a method derived by Astley (1990, 1991). A
dispersion relationship derived from a functional describing coupled fluid-structure
waves is used to compute the fundamental characteristics of the propagation of low
frequency (plane) waves within a duct. The result is further implemented into a two-port
formulation which is the main contribution of the present work.
The second part of this section deals with the radiation of sound from duct walls. Using
the equations derived for the internal sound propagation, one can analytically estimate
the level of sound emitted by the duct walls. The model used is based on a work by
Cummings (1978) on the so-called duct breakout effects. It regards the duct as a line
of monopole sources. This method gives good results as long as only the fundamental
(plane wave) propagation mode is investigated.
2.1 Governing equations, functional and dispersion relationship
A schematic outline of the duct with homogeneous properties along its z-axis is shown
in Figure 1.
where R is the cross-section area, C is the boundary, s is the length coordinate for the
flexible wall, c is the speed of sound (c0= 343 m.s1 in the airway), is the density and k
is the wavenumber. Waves propagating in lined flexible walled ducts are present in
three different media (wall, liner, air) with different acoustical characteristics. Acoustic
energy is transmitted along the duct in the form of coupled fluid-structured waves where
the wave energy is split between the fluid (air and liner) and the structure.
2.1.1
Air-borne waves
The acoustic pressure in the airway is assumed to be harmonic and has the following
form
p0 ( x , y , z ,t ) = p0 ( x , y )e i ( t z ) ,
(2-1)
where ^ denotes a complex valued (Fourier) quantity, z is the direction of the wave
propagation, i.e., along the axial direction of the duct and po is the pressure mode shape.
It is assumed that no airflow is present in the duct.
The wave equation governing the propagation of the first harmonic mode in the air
filled region (Region R0) is
0 + k 2 (1 2 )p0 = 0,
2 p
(2-2)
where the subscript 0 denotes the airway, k = c0 is the plane wave number,
The acoustic pressure in the liner is also assumed to be harmonic and is written as
pa = pa ( x , y )e i ( t t ) .
(2-3)
In the liner (Region Ra), the propagation of sound waves is characterised by the
following equation
a + (k a2 2 )pa = 0,
2 p
(2-4)
where the subscript a denotes the liner. ka is the frequency dependent liner complex
wavenumber which can be determined by standard models or experimentally using
various acoustical measurement techniques (cf Appendix).
2.1.3
Structure-borne waves
The flexible walls vibrate and radiate energy to the surrounding environment. This
structural vibration significantly influences the internal propagation of sound wave.
Assuming thin plate theory, the equation governing the normal displacement u of such
flexible walls is represented by
u (s , z ,t ) = u (s )e i ( t z ) ,
(2-5)
where s and z are the spatial coordinates for the flexible wall/plate.
Assuming only bending waves in the plate; the standard equation for a thin plate gives
2u
m 2 = B 4 u + pu ,
t
where m = s h , B = Eh
(12(1 ))
2
and k =
4
p
2m
B
(2-6)
. s is the plate density, h its
4 u
2 u
+ ( 4 k p4 )u = pu .
B 4 2
2
s
s
(2-7)
The pressure or dynamic loading applied on the plate (the pressure difference between
the upper and lower surface) is represented by pu (the subscript u denotes the flexible
wall) and can have different forms. However, the pressure along the wall is assumed to
be uniform in this paper due to the low frequency plane wave propagation.
Regarding the acoustical losses, three types exist:
-Structural internal loss factor
-Boundary loss factor
-Radiation loss factor
In the model they are comprised in only one single loss factor, the structural loss factor
(2-8)
where E0 is the standard Youngs modulus which can be found in any engineering hand
book. For example: E0 = 19.5 1010 Pa (Steel ) .
(' k )2 k p2
where k s =
pu
,
B (' 4 k p4 )
(2-9)
For rectangular flexible ducts with pairs of vibrating walls oriented along x and y, the
displacement fields are given by
u x (s ) = A cos (k s s ) + B sinh(k s s ) + C cos (k s s ) + D sin(k s s ),
(2-10)
(2-11)
(2-12)
u y ( 0 ) = u y ( S y ) = 0,
(2-13)
(2-14)
The uniform internal pressure assumption (plane wave range) implies that
the displacement for each plate has a slope equal to zero in the middle of the plate.
u x (S x 2 )
= 0,
s
u y (S y 2 )
s
= 0,
(2-15)
(2-16)
(2-17)
Boundary conditions
In order to express the problem in a variational form, one needs to state the different
boundary conditions of the problem.
At the interface between two media, the pressure p and the normal displacement should
be conserved, i.e., be continuous across the interface.
Using the appropriate momentum equation relating the pressure and the displacement, it
is possible to write down the different boundary conditions.
On C1 (along the rigid wall), the velocity in the direction of the normal is zero, it
implies that
pa n1 = 0.
(2-18)
On C2 (between the liner and the airway), there should be continuity in the particle
displacement along the normal, i.e.,
(p0 n2 ) =
a 2
(pa n2 ).
(2-19)
On C3 (along the plate), there should be continuity in the displacement in the normal
direction, i.e.,
1
a 2
2.1.5
pa n3 = u
(2-20)
From the wave equations ((2-2), (2-4) and (2-7) and boundary conditions ((2-18), (2-19)
and (2-20)) for this problem, it is now possible to create a single expression called a
functional. This functional is stationary for acoustic and vibration fields that satisfy the
equations and the boundary conditions. This functional is then used to setup finite
element schemes or to derive approximate solutions using suitable ansatz functions.
After some manipulations, it is found (Astley (1990)) that the functional F for this
problem is given by
2
1 2 u 2
2
2 u
F (u , p ) = 0 g 2 + 2k ' + (k 4' 4 k p4 )u 2 upu ds
2
s
C3
1
0 p0 k 2 (1 ' 2 )p02 }dxdy + 0
{
p
2 R0
2a
2.1.6
{p
(2-21)
Ra
Analytical formulation
The functional (2-21) forms the basis for a simple analytical analysis from which a
dispersion relationship is obtained. However, one has to make complementary
assumptions to further proceed. Generally, when the acoustic wavelength is larger than
the characteristic dimension of the duct, it may be suitable to see the wave as a plane
wave, that is, all acoustic properties such as pressure and velocity can be considered as
constant over the cross-section of the duct. Thus,
p ( x , y ) = pu = pa = p0 = pint = cons tan t
(2-22)
(2-23)
where pint represents the uniform pressure applied on the inner surface of the flexible
wall and u ' (s, ' ) is the solution of (2-7) for which pint is set to unity (unit pressure).
Incorporating this into (2-21) the functional F is rewritten in the form
2
0 2 k a
1
1 2 2
2
2
F = ik (' )Spint + k pint R0 {' 1}+ Ra
' ,
2
2
a
k
(2-24)
( ) = (1 S ) ds = (i 0 c0 S ) u ds
(2-25)
is the averaged wall admittance over the vibrating length S. a and ka correspond
respectively to the liner density and wavenumber. Minimisation of this functional with
respect to pint (plane wave approximation) gives the dispersion relationship (2-26) from
which the roots are sought
G (' ) =
2
1
1
k
ik (' )S + k 2 R0 {' 2 1}+ Ra 0 ' 2 a .
2
2
a
k
(2-26)
4 (m B ) .
The result that emerges is that the root obtained from this scheme converges to the same
coupled wave number for any initial guess. The reason for this is the ansatz in (2-23)
which will always track an acoustic type of mode, i.e., the one that transports acoustic
energy in the air. For the in-duct sound transmission problems of interest here, it will be
10
assumed that it is sufficient to only include this mode type in the analysis.
2.2 Flexible duct transfer matrix formulation
The predicted coupled propagating wavenumber is now used to build the 2-port
transfer matrix corresponding to the flexible duct element. This transfer matrix is
suitable for calculation of the element transmission loss index or can be inserted in an
acoustic network composed of various other elements for which the transfer matrices
are known. The block technique consists of multiplying the transfer matrices to obtain
the network transfer matrix.
The transfer matrix for an element is characterised by two independent state variables:
acoustic pressure p and volume velocity q on each side of the element.
Figure 2: Duct transfer matrix defining a relationship between the plane wave state variables
(p,q) at the input 1 and the output 2 cross sections.
(2-27)
Assuming a straight homogenous duct section along z with length l and low Mach
number (M<<1) (cf Figure 1) , the duct transfer matrix is (Munjal (1987), Davies(1988))
cos( l ) iZ d sin( l )
T11 T12
= i sin( l )
Td =
,
cos(
l
)
T21 T22 Z
d
(2-28)
T12
.
T21
(2-29)
11
flow and equal input and output impedance Z p = 0 c0 R where R is the cross-section
area of the pipe.
1
T
ITL = 10 log10 T11 + 12 + Z p T21 + T22
4
Zp
2.2.1
(2-30)
The duct transfer matrix Td formulated in (2-28) necessitates the coupled wavenumber
obtained after solving the dispersion relationship (2-26) and the duct internal impedance
Zd evaluated as follows
r
v
= p .
t
(2-31)
Thus, one gets in the absorber and in the airway for waves in positive direction
Absorber: a i va = p + ,int i ,
(2-32)
Airway: o i vo = p + ,int i ,
(2-33)
a va+ = 0 v0+ .
(2-34)
12
(2-35)
Zd =
p+ ,int
.
=
+
q
Ra R0
+
a 0
(2-36)
The same results with a change of sign will apply to negative direction.
2.3 The line source method for acoustic radiation/breakout from duct walls
0 q
(2-37)
where q is the volume velocity per unit length and can be expressed in terms of the
averaged wall admittance as
q =
pint
S .
0 c0
(2-38)
The averaged wall admittance depends on the geometry of the flexible wall and is
expressed in equation (2-25).
For a duct of finite length l, the radiated sound power for any type of waves is obtained
using a radiation efficiency factor Cr. Equation (2-37) then becomes
2
Wrad
C q
= W C r = r 0
.
8
(2-39)
Cr is obtained graphically from Figure 4 and is taken from Cummings (1978) where
13
k=/c0 and =k is the axial propagating wave obtained from the dispersion
relationship (2-26).
The line source model derived originally by Brown & Rennison (1974) only considered
one single travelling wave. In an attempt to refine this model, Cummings (1980)
incorporates an additional reflected wave. A modified radiation efficiency factor Cr is
thus derived from the assumption of two sinusoidal travelling waves. Cummings (1980)
claims that it does not affect the total radiated power. Therefore, the single wave
assumption for the radiation model is valid even when structural reflections occur.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-10
-8
-6
-4
-2
0
l(1-k/)
10
Concerning now the internal sound power in the positive direction W+,int, it is defined
from
1
2
W+ ,int = p + ,int Re ,
Zd
(2-40)
(2-41)
Equations (2-39) and (2-40) are used to derive the expression for the radiation
14
2
8 0 c0 Re[ (Ra a + R0 0 )]
RTL = 10 log10
2
.
C r 2 S 2
(2-42)
Theoretical models have been derived concerning the internal propagation and radiation
of plane waves in rectangular ducts with flexible walls. The models give the
transmission loss due to the internal propagation and the radiation (breakout effect).
Experimental evaluation is presented in the next section in order to validate the theory.
15
16
Types of ducts
The geometry of the first duct tested can be seen in Figure 5 and Figure 6. This duct has
been built at MWL and represents a first benchmark test for the theory. The idea was to
have a duct with a single flexible wall and with well defined boundary conditions.
Different duct types are tested. The first one, shown in Figure 6, consists of three
wooden rigid walls and a flexible wall made of steel (thickness: 0.5 mm).
Figure 5: Specification of the duct with a single flexible wall (length l=2m). The liner has a
thickness of 22 mm.
17
18
Figure 7: Variation of the boundary conditions at the inlet and outlet section, I : clamped, II:
unclamped.
19
20
Figure 9:Schematic of the experimental rig.
Corrugated duct
The duct with a single flexible wall is also used to obtain the liner properties (cf
Appendix for more details). It is necessary to remove the flexible plate and replaced it
with a thick wooden plate. The duct is then made completely rigid as shown in Figure
11.
Figure 11: Rigid walled duct. The polyurethane foam is lined at the bottom.
21
3.1.2
Electronic equipment
Sound waves are generated from two loudspeakers (Figure 9) connected to a Tektronix
signal processor which is used to produce the signal fed to the loudspeaker. The
microphone is connected to the second channel. Correlation between the input signal
and microphone response is performed. The spacing between the microphone locations
dictates the range of frequency under investigation. For a measurement frequency
ranging between 50 and 2000 Hz, a minimum spacing of 70 mm and of maximum
spacing of 270 mm are required.
3.2 Internal experimental measurement techniques
The experimental procedures used to measure the sound fields in the different duct
sections are now described. The separate measurement techniques to characterise the
liner properties are given in the appendix.
3.2.1 Experimental duct transfer matrix
The standard two-microphone technique (Bodn & bom (1986)) applied separately on
the upstream and downstream pipe is used to determine the duct transfer matrix Td of
the acoustical element tested. The procedure used is based on the so-called source
switching technique described for instance by bom (1999).
One can define the duct transfer matrix Td as
p1 T11 T12 p2
=
q1 T21 T22 q2
straight duct with constant
=
properties along its axis
cos( l ) iZ d sin( l ) p
2
i sin( l )
q ,
cos(
l
)
Z
2
d
(3-1)
where is the propagation wavenumber and Z d = T12 T21 is the internal duct
impedance.
Only one microphone is used at six different locations (Figure 9). Although it is time
consuming to move the microphone to the various locations, the necessary phase
calibration between multiple microphones is eliminated.
First measurements of the transfer functions between the loudspeaker voltage and one of
the six microphone positions are done using loudspeaker LS I (Figure 9).
22
The procedure is then repeated for the second loudspeaker LS II. A total of 12
measurement files are then collected during each measurement session. A Matlab code
is used to obtain the transfer matrix Td and internal transmission loss ITL (see (2-30))
3.2.2 Incident power and duct internal sound power
The test ducts are positioned between the two circular pipes. Measurement of transfer
functions at different microphone positions is carried out to decompose the waves as
shown in Figure 12.
23
(3-2)
where s is the distance between microphone positions 1 and 2. In matrix form this
becomes
1 p + ,upstream
.
e iks p ,upstream
p1 1
p = e iks
2
(3-3)
+ ,upstream and p
,upstream (wave decomposition) yields
Rearranging (3-3) to extract p
p e iks p2
,
p + ,upstream = 1
2i sin ks
(3-4)
p1e iks + p2
p ,upstream =
.
2
i
sin
ks
(3-5)
(3-6)
H e / 1e iks + H e / 2
.
2i sin ks
(3-7)
H e / p+ ,upstream =
H e / p ,upstream =
Recalling that
H e / p+ ,upstream
H e / p ,upstream
S p+ p+
S ee
=
S p p
S ee
(3-8)
,
(3-9)
+ ,upstream and
where S p p and S p+ p+ are the uncalibrated autospectra of the incident p
reflected wave p ,upstream and See the loudspeaker autospectrum.
The true autospectra for the travelling waves are given by
2
2
2
S tp+ p+ = K cal
S p+ p+ = K cal
H e / p+ ,upstream S ee ,
2
2
2
S t p p = K cal
S p p = K cal
H e / p ,upstream S ee ,
(3-10)
(3-11)
where Kcal is the calibration factor [Pa/V] of the quarter-inch microphone mounted on
the circular pipe.
24
+ ,int , p
,int ) and duct internal power W+,int
Travelling pressure waves inside the duct ( p
,int is small and neglected. The duct
Due to the reflection free termination of the duct, p
+ ,int is derived as
internal pressure p
+ ,int = Tp+ ,upstream ,
p
(3-12)
2
,
1 + Z p Zd
(3-13)
where Zp and Zd are the pipe and duct impedances respectively. The duct internal power
W+,int is calculated using (2-40).
3.3 Duct radiated sound power (external measurement)
Measurements are performed in the reverberation room at MWL using a Brel & Kjaer
reference source with calibrated third octave band frequency data. Conversion of these
data to smallband with the same frequency resolution as the one obtained from the one
used in the in-duct measurements
Lw nB = Lw1 3 10 log10 ( n ),
(3-14)
where Lw nB is the sound power level in narrowband, Lw1/3 is the sound power level in
1/3-octave band from the source calibration data sheet and n is the number of narrow
band points within a 1/3 octave band. White noise within each band is assumed. A
rotating half-inch microphone in the reverberation room at a distance of at least 1m
away of any reflecting surface is used to measure the sound radiated. A room factor K is
determined according to
K( f ) =
Wref source
S pp
(3-15)
where Wref source is the narrow band converted reference source calibration data and Spp is
the rotating microphone autospectrum. Once the room factor K has been obtained, the
sound power radiated per unit length by the duct of length l is given by
Wrad = KS pp (duct ) l .
(3-16)
The radiation transmission loss RTL defined in equation (2-41) can now be obtained by
combining the measured in-duct sound power Wint and radiated power Wrad.
25
26
= 1e
bw1 f f r 1
f r 1
+ 2 e
bw2 f f r 2
f r 2
(4-1)
= 1e
bw1 f f r 1
f r 1
+ 2 e
bw2 f f r 2
f r 2
+ 3e
bw3 f f r 3
f r 3
+ 4 e
bw4 f f r 4
f r 4
(4-2)
where fr corresponds to the resonance frequencies so that a maximum loss factor occurs
at these frequencies and bw corresponds to the bandwidth factor. These parameters are
set to the values tabulated in Table 1.
Duct
27
4.1
In-duct investigation
The first series of measurements are performed on a duct with three rigid walls and a
flexible plate made of steel (thickness: 0.5 mm). Measured and predicted internal
transmission loss results are shown in Figure 13.
45
Exp
Model
40
35
30
25
20
15
10
5
0
200
400
600
800
Frequency(Hz)
1000
1200
Figure 13: Predicted and measured internal transmission loss for the unlined rigid duct with one
flexible plate.
The predictions are in acceptable agreement with the measured data, the discrepancy
being approximatively 1 dB over the whole range frequency. The largest transverse
dimension for this duct is 169 mm yielding a cut-off frequency of 1000 Hz. This can be
seen in the measured plot with an increase of the number of peaks for frequencies above
this cut-off frequency. Concerning now the resonances in the flexible wall
corresponding to the peaks in the transmission loss, they occur twice within the
investigated frequency range. The first one is attributed to a frequency of 93 Hz and 510
Hz for the second resonance. It can be mentioned that it has been necessary to adjust the
value of the plate Youngs modulus used in the model so both measured and theoretical
resonance peaks would coincide. The value has to be increased by 40%. Inspection of
the plate revealed that due to its small thickness (only 0.5 mm) it has been buckled and
it is not flat when mounted.
28
Internal sound propagation for the lined duct with one flexible wall
Theoretical and measured internal transmission loss results are illustrated in Figure 14
and Figure 15 for the case when the polyurethane foam sheet is lined with no air gap.
In order to achieve this experimentally, the steel plate is taken off the duct and the liner
piece is put in place so part of it comes off the top of the duct. The liner is then pushed
evenly downwards with the aid of the flexible wall to ensure no air gap between the
wall and the liner. The effect of lining the flexible wall with polyurethane foam on the
measured internal transmission loss is seen in Figure 14.
40
Unlined
Lined
35
30
One flexible walled duct
25
20
15
10
5
0
200
400
600
800
Frequency (Hz)
1000
1200
Figure 14: Experimental transmission loss for the one walled flexible duct.
45
Exp
Model
40
35
30
25
20
15
10
5
0
200
400
600
800
Frequency(Hz)
1000
1200
Figure 15: Predicted and measured internal transmission loss for the lined rigid duct with one
flexible plate.
As seen from Figure 15, the presence of the liner does not shift the two resonance
frequencies; however, the magnitudes are reduced. The first resonance occurring at a
29
frequency of 93 Hz is still the dominant peak. As expected lining the duct, even on only
one wall improves drastically the internal transmission loss at high frequencies.
4.1.2
The transmission loss result for an empty duct is shown in Figure 16.
40
Exp
Model
35
30
25
20
15
10
5
0
200
400
600
800
Frequency(Hz)
1000
1200
Figure 16: Predicted and measured internal transmission loss for the unlined corrugated duct.
Due to the fact that two pairs of walls of different lengths (150 and 200 mm) are
excited, four resonance peaks are seen in Figure 16. A further analysis shows that these
resonances occur at the following frequencies
Corrugated walls
150 mm
82 Hz, 640 Hz
200 mm
30
The porous material used previously in the first duct is now put on the bottom wall of
the corrugated duct. As for the duct with one flexible wall (Figure 14), lining the
corrugated duct modified the measured internal transmission loss curve in a similar
fashion (Figure 17). Experimental and theoretical data for this lined corrugated duct are
shown in Figure 18.
25
Unlined
Lined
Rectangular four walled flexible duct
20
15
10
5
0
200
400
600
800
1000
Frequency (Hz)
1200
Figure 17: Experimental internal transmission loss for the lined corrugated duct.
31
30
Model
Exp
25
Rectangular four walled flexible duct
20
15
10
100
200
300
400
800
900
Figure 18: Internal transmission loss for the lined corrugated duct.
Comparison of Figure 15 (lined duct with one flexible wall) and Figure 18 (lined
corrugated duct) shows, except at the peaks, transmission loss values that are more or
less equal. In practice, this implies that the effect of flexible walls in a lined duct is only
of importance around structural resonances, in particular the first.
4.1.3
Measurements on the liner mounted in a rigid walled duct have also been performed to
determine the liner properties, see appendix A. From these results, the transmission loss
for a lined rigid walled duct can be obtained, see Figure 19.
30
25
20
15
10
0
100
200
400
600
800
Frequency(Hz)
1000
1200
Figure 19: Measured internal transmission loss for the lined completely rigid duct.
32
In Figure 20, the transmission loss for all the three tested lined ducts is presented. As
can be seen the differences are small except at the first resonance peak.
30
Rigid duct
Rectangular flexible duct
First duct (one flexible wall)
25
20
15
10
100
200
300
400
800
900
Figure 20: Comparison of measured internal transmission loss for the three lined ducts
investigated.
The sound radiation from the ducts is measured and theoretically evaluated in the
following section. Due to the uncertainties of the dynamic properties of the duct with a
single flexible wall, only the sound radiation results for the corrugated are presented.
Sound radiation transmission loss for the corrugated unlined duct
Measurement and theoretical radiation transmission loss results for the unlined
corrugated duct are shown in Figure 21 in narrow band with the duct impedance Zd
determined experimentally (see (2-29)) or theoretically (see (2-36)) and in Figure 22 in
1/3 octave band.
33
45
Experimental
Model+Theoretical duct impedance Z d
Unlined
External radiation transmission loss (dB)
40
35
30
25
20
15
10
5
100
200
300
400
500
600
700
800
Frequency (Hz)
900
1000
1100
1200
Figure 21: Radiation transmission loss for the unlined corrugated duct.
60
Experiment
Theory
Unlined
55
50
45
40
35
30
125
160
200
250
315
400
500
Frequency (Hz)
630
800
1000
Figure 22: 1/3 octave band radiation transmission loss for the unlined corrugated duct.
The effect of lining the bottom of the corrugated duct on the radiation transmission loss
is shown in Figure 23 and Figure 24.
34
45
Experimental
Model+Theoretical duct impedance Z d
lined
40
35
30
25
20
15
100
200
300
400
500
600
700
Frequency (Hz)
800
900
1000
1100
Figure 23: Predicted and measured radiation transmission loss for the lined corrugated duct.
65
Experiment
Theory
60
Lined
55
50
45
40
35
30
125
160
200
250
315
400
500
Frequency (Hz)
630
800
1000
Figure 24: 1/3 octave band radiation transmission loss for the lined corrugated duct.
The effects of some parameters on the interior sound transmission are now investigated.
4.3.1
The effect of the position of the liner material inside the duct with one flexible plate is
now studied. Measurement of the transmission loss is performed and shown in Figure 25
for two different lining methods.
The liner is positioned at the flexible wall and at the opposite side against the rigid wall.
In both cases there is no air gap between the liner and the wall.
35
45
40
35
30
25
20
15
10
5
0
-5
100
200
300
400
800
900
Figure 25: Influence of the lining position within the one walled flexible duct.
As can be seen from Figure 25 the position of the liner is not critical.
4.3.2
Structural termination effects on the internal transmission loss for the first duct (single
flexible wall) are studied and experimental results are given in Figure 26. Two
termination conditions are studied: clamped and unclamped (Figure 7). The coupling
between the acoustic and the structural wave is mainly important around the structural
resonances. This is clear for instance from comparing the transmission loss for the lined
flexible duct and the lined rigid duct in Figure 20. Also in Figure 26 one can see that the
termination conditions affect the behaviour around the structural resonances.
36
40
Clamped
Silicone sealing (unclamped)
35
30
25
20
15
10
5
0
100
200
300
400
800
900
Figure 26: Internal transmission loss in the one walled unlined flexible duct with different
experimental axial boundary conditions.
For a thorough treatment of the radiation to the surrounding, the calculation of the total
internal sound pressure pint should include the reflected internal sound pressure waves
,int as
p
int ( z , t ) = p+ ,int e i ( t z ) + p
,int e i ( t + z ) .
p
(4-3)
That leads to
pint
(4-4)
1
int
= p
l0
(z )dz = p+ ,int
,int +
+ p
+ ,int p
,int
p
2kl
sin( 2kl ),
(4-5)
where l is the axial length of the duct. As long as kl>>1 the cross term can be neglected.
Maximum reflectivity (R=1) at the outlet of the duct implies then that
+ ,int = p ,int leading to a maximum internal duct power increase of 3 dB.
p
37
38
5. Conclusions
In this paper low frequency (plane wave) transmission in lined ducts with flexible walls
has been studied.
These models have been compared with experimental data and a satisfactory agreement
has been obtained, especially for an industrial corrugated duct. The results are in
accordance with the findings of Astley (1990) concerning the internal propagation and
Cummings (1980) for the radiation. Peaks and dips attributed to resonances in the walls
are well predicted.
The main contribution in this work is the implementation of the results in 2-port or
transfer matrix form. The model can be implemented in the software code SID
(developed at MWL) or in other codes based on this formulation.
39
40
6. References
1. bom M., 1999, Determination of porous material data via two-port measurements,
Proceedings of Inter-Noise 99.
2. Almgren M, 1982, Prediction of sound pressure level outside a closed ventilation duct a
literature survey, Report F82-03, Chalmers University of Technology, Department of Building
Acoustics.
3. ASHRAE, 2003, ASHRAE HVAC Applications Handbook, chapter 47: Sound and Vibration
Control
4. Astley, R. J., 1990, Acoustical modes in lined ducts with flexible walls. A variational
approach, Proceedings of Inter-Noise 90, 575-578.
5. Astley R. J., 1991, A finite element scheme for acoustic propagation in flexible-walled ducts
with bulk-reacting liners, and comparison with experiment, Journal of Sound and Vibration,
150(1), 119-138.
6. Astley R. J., Cummings A., 1984, A finite element scheme for acoustic transmission through
the walls of rectangular ducts: comparison with experiment, Journal of Sound and Vibration,
92(3), 387-409.
7. Astley R.J., Cummings A., 1987, A finite element scheme for attenuation in ducts lined with
porous material: comparison with experiment, Journal of Sound and Vibration, 116(2), 239263.
8. Astley R.J., Cummings A., Sormaz N., 1991, A finite element scheme for acoustical
propagation in flexible walled ducts with bulk reacting liners, and comparison with experiment,
Journal of Sound and Vibration, 150, 119-138.
9. Bodn H., bom M., 1986, Influence of errors on the two-microphone method for measuring
acoustic properties in ducts, Journal of the Acoustical Society of America, 79(2), 541-549.
10. Bodn H., Carlsson U., Glav R., Wallin H. P., bom M., 2001, Ljud och vibration,
Norstedts Tryckeri AB, Stockholm.
41
11. Brown G. L., Rennison D. C., 1974, Sound radiation from pipes excited by plane acoustic
waves, Noise, Shock & Vibration Conference, Monash University, Melbourne.
12. Cabelli A., 1985a, The propagation of sound in a square duct with a non-rigid side wall,
Journal of Sound and Vibration, 103(3), 379-394.
13. Cabelli A., 1985b, Application of the time dependent finite difference theory to the study of
sound and vibration interactions in ducts, Journal of Sound and Vibration, 103, 13-23.
14. Cummings, A., 1976, Sound attenuation in ducts lined on two opposite walls with porous
material, with some applications to splitters, Journal of Sound and Vibration, 49(1), 9-35.
15. Cummings, A., 1978, Low frequency acoustic transmission through the walls of rectangular
ducts, Journal of Sound and Vibration, 61(3), 327-345.
16. Cummings, A., 1979a, Low frequency sound transmission through the walls of rectangular
ducts: further comments, Journal of Sound and Vibration, 63(3), 463-465.
17. Cummings, A., 1979b, The effects of external lagging on low frequency sound transmission
through the walls of rectangular ducts, Journal of Sound and Vibration, 67(2), 187-201.
18. Cummings, A., 1980, Low frequency acoustic radiation from duct walls, Journal of Sound
and Vibration, 71(2), 201-226.
19. Cummings, A., 1981a, Design charts for low frequency acoustic transmission through the
walls of rectangular ducts, Journal of Sound and Vibration, 78, 269-289.
20. Cummings, A., 1981b, Stiffness control of low frequency acoustic transmission through the
walls of rectangular ducts. Journal of Sound and Vibration, 74, 351-380.
21. Cummings, A. 1983a, Approximate asymptotic solutions for acoustic transmission through
the walls of rectangular ducts, Journal of Sound and Vibration, 90(2), 211-227.
22. Cummings, A., 1983b, Higher order mode acoustic transmission through the walls of
rectangular ducts, Journal of Sound and Vibration, 90(2), 193-209.
23. Cummings, A. 1994, The attenuation of sound in unlined ducts with flexible walls, Journal
of Sound and Vibration, 174(4),
42
24. Cummings, A., 2001, Sound transmission through duct walls, Journal of Sound and
Vibration, 239(4), 731-765.
25. Cummings A., Astley R.J., 1995, The effects of flanking transmission on sound attenuation
in lined ducts, Journal of Sound and Vibration, 179, 617-646.
26. Davies P. O. A. L., 1988, Practical flow duct acoustics, Journal of Sound and Vibration,
124(1), 91-115.
27. Ducret F, bom M., 2003, Low frequency sound propagation in rectangular ducts, Tenth
International Congress on Sound and Vibration, Stockholm, Sweden.
28. Elnady T., 2002, On the use of liners to reduce aircraft jet engine noise, Licenciate Thesis,
Royal Institute of Technology, Department of Aeronautical and Vehicle engineering, TRITAFKT 2002:26.
29. Glav R., 1994, On acoustic modelling of silencers, Doctoral thesis, Royal Institute of
Technology, Department of Aeronautical and Vehicle engineering, TRITA-FKT Report 9435.
30. Kinsler L. E., Frey A. R., Coppens A. B., Sanders J. V., 1982, Fundamentals of acoustics,
John Wiley & Sons, Inc.
31. Kirby R., Cummings A., 1998, Structural/acoustic interaction in air-conditioning ducts in
the presence of mean flow, Proceedings of ISMA23, Leuven, Belgium, 677-684.
32. Martin V., 1991, Perturbation of fluid-guided waves introduced by bending plates, Journal
of Sound and Vibration, 144, 331-353.
33. Martin V., Cummings A., Gronier C., 2004, Discrimination of coupled structural/acoustic
duct modes by active control: principles and experiments results, Journal of Sound and
Vibration, 274, 583-603.
34. Munjal M.L., 1987, Acoustics of ducts and mufflers, Wiley Interscience, New York.
35. Nygrd S, 2000, Modelling of low frequency sound in duct networks, Licenciate Thesis,
Royal Institute of Technology, Department of Aeronautical and Vehicle engineering, TRITAFKT 2000:57.
43
44
The sample is mounted at the end of tube as shown in the following picture.
(A-1)
iZ a sin (k a t ) p2
,
cos (k a t ) q2
45
(A-2)
where Za=aca/Ra .
This leads to
Z1 =
p1
cos (k a t )Z 2 + iZ a sin(k a t )
=
,
q1 (i sin (k a t ) / Z a )Z 2 + cos (k a t )
(A-3)
where Z 2 = p2 q2 . Za and a are thus determined from the measurement of the normal
incident impedance Z1 for samples of two different thickness t.
For an open end termination, three measurements are necessary:
-Measurement on an empty duct (no sample) for determination of Z2 (radiation
impedance)
-Measurement of Z1,sample
and Z1,sample
thicknesses t1 and t2
It follows that
Z1,sample 1 (i sin (k a t1 )Z 2 + Z a cos (k a t1 )) Z a cos (k a t1 )Z 2 + iZ a2 sin (k a t1 ) = 0 ,
(A-4)
(A-5)
Equations (A-4) and (A-5) can be solved iteratively using a Newton-Raphson scheme to
obtain Za and ka=/ca.
Two-port measurement (two-sided technique)
The sample is mounted in a rigid pipe. Cautions have to been taken to ensure no air gap
between the wall of the pipe and the sample.
Measurement on each side of the sample using the two-microphone technique allows
determination of the sample transfer matrix T (bom (1999))
46
T12 p2 cos (k a t )
p1
p T
= [T ] 2 = 11
=
q1
q2 T21 T22 q2 i sin(k a t ) / Z a
iZ a sin (k a t ) p2
.
cos (k a t ) q2
(A-6)
Using the elements of the sample transfer matrix T, the sample properties can be derived
as
Za =
a ca
Ra
T12
,
T21
1
ln (T11 + T22 + T12 / Z a + T21 Z a )
2
.
ka =
=
ca
it
(A-7)
(A-8)
Rigid duct
Here the liner is mounted similar to the test cases but in a rigid walled duct as shown in
Figure 27.
q .
cos(
l
)
q1
q2 T21 T22 q2 Z
2
d
47
(A-9)
Using this measured transfer matrix and assuming no reflected wave, it yields the
experimental duct impedance Zd expressed as (see (2-36))
Zd =
+
p
=
q+
Ra
Ro
T12
.
T21
(A-10)
a =
Z d 0 Ra
.
0 R0 Z d
(A-11)
The complex axial wavenumber is obtained from the elements T11 and T12 in the
measured transfer matrix Td using (A-8).
For a rigid walled duct with no mean flow, Astley (1990) expresses the propagating
coupled wavenumber as
Ra o ka2 + Ro a k 2
Ra o + Ro a
(A-12)
R0 ( 2 k 2 ) a + Ra 0 2
.
Ra 0
48
(A-13)
Table of Contents
1. Introduction .................................................................................................................. 1
1.1
Earlier related work .......................................................................................... 2
1.1.1
Propagation in straight pipes .................................................................... 2
1.1.2
Radiation from openings .......................................................................... 2
1.1.3
Flow generated noise ................................................................................ 2
1.2
Pressure reflection coefficient and impedance at openings.............................. 5
1.3
Description of the airflow facility and two-microphone technique.................. 6
2. Acoustics of a straight pipe .......................................................................................... 9
2.1
Methods and materials...................................................................................... 9
2.1.1
Presentation of the straight pipe rig.......................................................... 9
2.1.2
Damping of plane wave.......................................................................... 13
2.1.3
End correction ........................................................................................ 13
2.1.4
Sound absorption by vorticity shedding in the near field....................... 14
2.2
Results and discussion .................................................................................... 17
2.2.1
Internal damping in the boundary layer of a straight pipe...................... 17
2.2.2
End correction ........................................................................................ 21
2.2.3
Sound absorption due to vorticity shedding ........................................... 23
2.3
Summary......................................................................................................... 26
3. Acoustics of tailpipes ................................................................................................. 27
3.1
Methods and materials.................................................................................... 27
3.1.1
Presentation of the tailpipe rig................................................................ 27
3.1.2
Presentation of the tailpipe elements ...................................................... 29
3.1.3
Modelling of pressure reflection coefficients and impedances .............. 33
3.1.4
Experimental procedures ........................................................................ 36
3.2
Results and discussion .................................................................................... 36
3.2.1
Reflection coefficient and impedance .................................................... 36
3.3
Summary......................................................................................................... 64
4. Modelling of exhaust lines ......................................................................................... 65
4.1
Description of an exhaust line ........................................................................ 65
4.2
The automotive exhaust line........................................................................... 67
4.3
The single pipe exhaust line ........................................................................... 71
4.4
Summary......................................................................................................... 74
5. Flow noise for a straight pipe and tailpipes................................................................ 75
5.1
Methods and materials.................................................................................... 75
5.1.1
Measurement of flow-induced noise in a reverberant room................... 75
5.1.2
Flow noise generation for a straight pipe ............................................... 76
5.1.3
Scaling laws............................................................................................ 78
5.2
Results and discussion .................................................................................... 82
5.2.1
Scaling law results for a straight pipe..................................................... 82
5.2.2
Flow-induced noise and scaling law results for tailpipes ....................... 82
5.2.3
Scaling laws review ................................................................................ 89
5.3
Summary......................................................................................................... 92
6. Summary and conclusions .......................................................................................... 93
7. References .................................................................................................................. 97
1. Introduction
A typical modern exhaust system is composed of one or several mufflers, a device for
exhaust gas cleaning and an exhaust pipe also called tailpipe. Exhaust pipes are
inevitably associated with cars, trucks and motorcycles. They can also be found in gas
turbine and jet engine systems. Exhaust noise is an important source of noise from an
internal combustion (IC) engine and effective exhaust mufflers are therefore needed.
Exhaust noise can also radiate mainly from the shell of expansion chamber mufflers.
Exhaustive research has been conducted on mufflers to find efficient types that with a
minimum pressure drop can sufficiently reduce the pressure pulsations emitted from the
engine manifold.
Mufflers are also to be found also in heating, ventilation and air-conditioning systems.
Two types of muffler exist. Reactive mufflers based on impedance mismatching that
create reflection of sound; this is usually done by a sudden change of geometry such as
an expansion chamber. Dissipative mufflers are characterised by the presence of porous
material or flow constrictions. Acoustical energy is dissipated through visco-thermal
effects in the porous material or via turbulence dissipation. Usually, both of these
mechanisms are present in a muffler. The impedance mismatch is used for low
frequencies (tones) and the dissipative mechanism at mid and high frequencies. For high
flow speeds, care must be taken so that the aeroacoustic noise produced by the muffler
itself does not become a problem. Mufflers have to be compact to fit , e.g., under a car.
In this paper the aeroacoustics of different exhaust tailpipes is investigated. Both the
acoustic impedance and the flow generated noise are investigated experimentally.
We start by investigating the different regions of a straight pipe with open end
termination. This study aims to set the foundations for another investigation concerning
the effects of changing the termination geometry (oblique cuts, diffusers and bends).
1.1
predict the flow noise sound power level for a specific configuration and flow speed.
Nelson & Morfey (1981) regard the flow separation process as a dipole type of source.
In other words, the physical mechanism of flow noise sound production can be
summarized as a fluctuating force acting at a specific position in the element. A similar
investigation on flow noise has been performed earlier for other geometries by Girath
and al. (2001) and Nygrd (2000).
The present document consists of four main parts. The first one concerns only the
straight pipe for which open-end reflection coefficient and impedance are treated
theoretically and experimentally under the variation of the flow Mach number. The
damping of the sound wave inside the pipe, the end correction and the absorption by
vorticity are also discussed.
The second part deals with tailpipes of various geometries (oblique cuts, diffusers and
bends). Only one aspect is considered for the tailpipe investigation: the sound pressure
reflection coefficient and impedance at the inlet and outlet. Measured low frequency
diffuser termination impedances are used in section three for calculation of the
performance of an exhaust line terminated by a diffuser. This shows the influence of
diffusing geometries and the flow Mach number on the efficiency to attenuate sound at
low frequency in an exhaust line. The fourth section concerns the sound power levels of
flow generated noise and presents scaling laws for the straight pipe and tailpipes. The
methodology systematically undertaken in this section is the comparison of the flow
noise produced by the tailpipes with the straight-pipe (reference case) results.
1. General methodology
This section describes the experimental rigs and theoretical analysis used for the
investigation of the acoustics of a straight pipe and tailpipes. The flow noise generated
from a straight pipe and tailpipes is treated in a separate chapter.
For both rigs, a rotating microphone is located in a reverberant room to measure the
flow noise. An additional measurement was performed using the rotating microphone
for the first straight pipe; the radiation of the sound from the straight pipe rig. It is
required to estimate the sound absorption in the vorticity.
1.2
1.3
Air flow was provided from the flow facility located on the KTH/MWL service floor ; a
maximum flow rate of 10 m3/s and pressure up to 10 kPa can be achieved. The flow is
taken to an anechoic room acting as a settling chamber. The pressurised anechoic room
then creates a flow in the test duct mounted between this room and the reverberation
room (see Figure 2).
A smooth inlet cone situated in the anechoic room is connected to the inlet of the rig to
ensure minimum flow background noise. Preliminary tests showed dips in the
measurement of the coherence function at the microphones due to resonances in the
pipe. In order to reduce these standing waves in the rig, an ordinary car muffler is
inserted upstream in the rig. Sufficient levels of excitation during measurement at
higher flow speeds is provided by two loudspeakers mounted onto the pipe further
downstream.
The airflow Mach number M measurement procedure consists of a Pitot tube connected
to a pressure transducer to ensure measurements of flow speeds in the pipe. The Pitot
tube is aligned with the centreline of the pipe where measurement of maximum
velocities is performed. For 4 103 Re 3.2 106 , the ratio of the mean to the
maximum velocity is U mean U max = 0.8 (Schlichting (1968)). The flow Mach number M
and wavenumber k+/- are assumed to be independent of the axial position (Peters and al.
(1993)).
For the plane wave decomposition in the pipes, the two microphone technique described
in bom & Bden (1988) is used. It is based on the transfer function between two flush
mounted microphones. The reflection coefficient and normal impedance can then be
calculated and moved to a specific position in the pipe.
The normalised1 acoustic impedance Z is calculated from the complex reflection
coefficient R using
Z = (1 + R ) (1 R ) .
(1-1)
In this paper Z refers to both the specific and normalised acoustic impedance. This is not completely
correct, the normalised impedance should be denoted as as in paper III.
f min =
0.05c0
0.4c0
,
f f max =
s
s
(1-2)
(1-3)
where k=/c0, M is the Mach number and l is the distance separating the reference
microphone to the straight pipe open end or to the tailpipe inlet as shown in Figure 1.
The acoustical damping over the length l is neglected.
The rig as shown in Figure 2 was constructed entirely of a single straight steel pipe of
length of 6 m, thickness of 2 mm and inner diameter of 35 mm (cut-off frequency of
5800 Hz). Six microphones are flushed mounted and are arranged in two clusters.
Cluster I is composed of microphones 1, 2 and 3 and cluster II corresponds to
microphones 4, 5 and 6.
The two-microphone technique frequency criteria (see equation (1-2)) implies
Cluster I
Cluster II
Microphone 1
Microphone 1
Microphone 6
Microphone 6
and 2, s12=28 mm
and 3, s13=218 mm
and 5, s65=28 mm
and 4, s64=173 mm
610 f 4900
80 f 640
610 f 4900
100 f 800
Twomicrophone
frequency
criteria (Hz)
The measurement is performed for various flow speeds in the pipe measured with a
Pitot-tube probe and is tabulated
Pitot-tube reading (m/s) (maximum
18
32
47
61
76
10
Figure 2: Layout of the straight pipe rig.
Cluster technique
(2-1)
5 = p6+ e ik+ ( x5 x6 ) + p6 e ik ( x5 x6 ) ,
p
(2-2)
4 = p6+ e ik+ ( x4 x6 ) + p6 e ik ( x4 x6 ) ,
p
(2-3)
where k+/- are the wavenumbers in the positive and negative directions.
Evaluation in matrix form gives
e ik ( x5 x6 )
p6+
1
= ik ( x x )
p e 5 6 e ik+ ( x5 x6 ) e ik+ ( x5 x6 )
1 p6
,
1 p5
(2-4)
(2-5)
which leads to
p6+ =
p6 =
ik ( x5 x6 )
ik ( x5 x6 )
1
e ik ( x5 x6 ) p6 p5 ,
e ik+ ( x5 x6 )
1
e ik+ ( x5 x6 ) p6 + p5 .
ik + ( x5 x6 )
e
(2-6)
(2-7)
To minimise the effect of noise on the measured data, equation (2-7) can be written in
terms of transfer functions H en = en e where en is the output voltage from microphone
n and e is the voltage to the loudspeaker in the test rig.
Equations ((2-5), (2-6), (2-7) can be rewritten in terms of these measured transfer
functions
11
H ee6 =
H ee =
6
where
ik ( x5 x6 )
ik ( x5 x6 )
1
e ik ( x5 x6 ) H e 6 H e5 / H 65c ,
ik + ( x5 x6 )
e
1
e ik+ ( x5 x6 ) H e 6 + H e 5 / H 65c ,
e ik+ ( x5 x6 )
(2-8)
(2-9)
which the two microphones are exposed to the same acoustic pressure.
Equations (2-8) and (2-9) imply that the pressure reflection coefficient at microphone 6
is
R6 = H ee / H ee+ .
6
(2-10)
Using the same procedure with the microphone pair 4 and 6 a lower frequency range is
covered.
The procedure is repeated for cluster I composed of microphones 3, 2, 1(reference) to
obtain the pressure reflection R1 at microphone 1.
Referring to cluster I, one derives for microphone 1 and 2
H ee + =
1
H ee =
1
e
e
ik ( x2 x1 )
ik ( x2 x1 )
(2-11)
(2-12)
1
e ik ( x2 x1 ) H e1 H e 2 / H 12 c ,
e ik+ ( x2 x1 )
1
e ik+ ( x2 x1 ) H e1 + H e 2 / H 12 c ,
ik + ( x2 x1 )
e
R1 = H ee / H ee+ .
1
(2-13)
The cluster technique can be used to determine the damping in the pipe. Two
formulations can be used. The first one (performed by the author) uses the travelling
pressure to determine the complex wavenumber whereas the second method (derived by
Peters and al. (1993)) uses the pressure reflection coefficient. The damping technique
consists essentially in combining the results obtained separately from the two clusters of
microphones.
It can be demonstrated that the relations between the travelling pressures at microphone
1 and 6 can be written as
(
= (H
)
)H
(2-14)
e ik ( x1 x6 ) = p1 / p6
(2-15)
ee1
/ H ee
6
16 c
Experimentally derived wavenumbers k+/- are extracted from (2-14) and (2-15) and
reinserted into the two-microphone technique ((2-8) and (2-11),(2-9) and (2-12)) for
12
each cluster iteratively until convergence is obtained. As starting values for the iteration
technique the classical no flow value for the wavenumber are used, i.e., k + = c0 . A
damping coefficient +/- is finally obtained from the absolute value of the imaginary
part of k+/-. An improved value for the complex reflection coefficient is thus obtained
from this method.
2.1.2 Damping of plane wave
Damping modelling
The interior losses in the fluid can normally be neglected (Pierce (1989)). Kirchoff
(1868) derived a formulation (2-16) for the visco-thermal damping that occurs in the
boundary layer in a straight pipe without flow. The second term in (2-16) proportional
to (1/Sh2) was added by Ronneberger (1975) and Tijdeman (1975). The attenuation of
acoustical energy is expressed as the imaginary part ( = Im(k ) ) of the wave number
kno flow expressed as
k no flow =
c0
1+
1
1 1 1
i
1 i 1
= i ,
1 + 1 / 2 2 1 + 1 / 2
Pr Sh
Pr
2 Pr c0
2 Sh
(2-16)
of radius a, 0 is the density and is the shear viscosity coefficient, Pr= Cp is the
Prandtl number, Cp is the specific heat coefficient and is the thermal diffusivity.
The modified Kirchoffs formulation suggested by Davies (1988) for the acoustic
wavenumber k+ and k- propagating respectively with and against the flow is expressed
as
k+ =
k no flow
1+ M
k =
k no flow
1 M
(2-17)
The end correction corresponds to the extended length required to obtain a phase shift of
180 between the incident and reflected wave. Its value normalised with respect of the
pipe radius is influenced by the opening geometry (Peters and al. (1993)) and by flow
13
(Boij (2003)). The low and high Strouhal limits of the magnitude of the end correction
have been found experimentally and theoretically to not be modified under the variation
of flow Mach number M. However, for intermediate Strouhal number around the
vicinity of St=1, the end correction behaviour is changed for different flow conditions
due to the strong coupling flow-acoustic fields in this region.
End correction determination
The end correction can be defined as the extra length that must be added to the pipe to
produce a reflection coefficient of -1. This implies
2 k
1 M 2
R = R0 e i
= R0 e i ,
(2-18)
( )(1 M 2 ) ,
2ka
(2-19)
where is the phase of the reflection coefficient R obtained from theoretically Munts
model (1977, 1990) or experimentally from the cluster technique and a is the pipe
radius.
Another way to derive the end correction is through the imaginary part of the
termination impedance Z,
Z=
p
= RZ + jX Z .
q
(2-20)
(2-21)
This section looks at the sound dissipation through vorticity. Flow separation at the lips
of the pipe (vorticity shedding) is known to absorb the acoustic energy (Bechert (1980),
Howe (1980). The sound wave energy is converted into vortex kinetic energy. The
14
vortices are propagating further downstream before they break up into turbulence and
converted into heat. An analytical model derived by Bechert (1980) is reviewed and
used to predict the losses due to vorticity shedding. An experimental method is
presented to obtain these losses. It requires the separate measurement of the incident and
radiated sound energies.
Analytical formulation for the sound absorption in the vorticity shedding
(2-22)
Wt = (2a 2 0 c0 )M p1+ .
2
(2-23)
It follows that the ratio of radiated sound power Wr to the transmitted sound power Wt is
Wr Wt = (1 + 43 M 2 ) 4 M (a c0 ) .
2
(2-24)
Measurement of the losses through vorticity requires the determination of the incident
power in order to estimate the acoustic power coming out of the pipe (transmitted
power). A separate measurement in a reverberation room allows the radiated acoustic
powers to be measured. The losses are obtained by comparison of the transmitted and
radiated powers.
In order to obtain experimentally the incident wave amplitude p+ at the opening, an
induct measurement is performed using the microphone cluster located next to the
opening (cluster I, cf Figure 2).
Using the definition of the microphone calibration factor K1 [V/Pa]
2
2
S e+ e+ = (K1 ) p1+ = H ee+ S ee ,
2
1 1
(2-25)
where S e+e+ is the autospectrum for the incident pressure expressed as V2, See is the
1 1
15
p1+
Wi =
0 c0
S pipe (1 + M ) =
2
H ee+ S ee
1
(K 1 ) 0 c0
2
S pipe (1 + M ) .
2
(2-26)
The incident power Wi can be combined with the pressure reflection coefficient R1
(equation (2-13)) to determine the sound power Wt transmitted out of the pipe
Wt =
p1+
0 c0
S pipe (1 + M ) R1 (1 M ) .
2
(2-27)
The radiated sound power level Wr has been measured in a reverberation room by
measuring the autospectrum from a rotating microphone and using a calibrated sound
source (ISO 3747).
The noise due to the flow generation in the pipe and at the opening has been measured.
It has been intended to compare this quantity to the sound power level transmitted from
the loudspeaker to the room to ensure a good signal-to-noise ratio. The measurements
are conducted using a broadband type of excitation as stipulated in standards for
measurement in a reverberation room (ISO 3741).
The experimental ratio of radiated sound power Wr to transmitted sound power Wt is
evaluated as
Wr
=
Wt
p1+
Wr
2
0 c0
S pipe (1 + M ) R1 (1 M )
2
16
.
(2-28)
2.2
A mean value Mach number M and Kirchoff damping are explicitly assumed if not
otherwise stated.
2.2.1 Internal damping in the boundary layer of a straight pipe
The influence of flow on acoustic damping can be seen in Figure 3. The theoretical
damping (see (2-16) and (2-17)) presented in Figure 3 corresponds to the absolute
value of the imaginary part of the calculated wavenumber k+/-.
0.16
0.14
No flow case
Damping (1/m)
0.12
0.1
0.08
0.06
+ (Exp)
0.04
- (Exp)
0.02
0
50
Theory
500
3000
3500
4000
+ (Exp)
0.25
Damping (1/m)
1000
- (Exp)
+ (Theory)
0.2
- (Theory)
0.15
0.1
0.05
0
50
M=0.14
500
1500
2500
Frequency (Hz)
3500
4500
17
The experimental results are obtained using the two-microphone and cluster techniques
and with formulation based on travelling pressures (cf (2-14), (2-15)). The cluster
technique consists of an array of six in-duct flushed mounted microphone.
The experimental results for the no flow case in Figure 3(a) agree well with the
calculated results using the Kirchoff formulation (see (2-16)). No significant difference
in damping is seen between the upstream and downstream propagating waves.
Comparison of the plots in Figure 3 shows that introducing flow into the pipe does not
significantly change the damping + of the downstream propagating wave. An increase
in the damping is however observed for the upstream propagating wave - (reflected
wave). It also appears that Davies modified theory (Davies 1988) of Kirchoff for the
case with flow (see equation (2-17)) underpredicts the damping.
A transmission loss TL [dB/m] due to the viscous-damping can be calculated from the
value of the damping [1/m] measured and shown in Figure 3. The TL is given as
TL = 8.7 .
(2-29)
Using the maximum value (=0.2) observed at 4.5 KHz for the flow case M=0.14 gives
a TL of 2 dB/m.
This result shows that attenuation through viscous boundary layer along the inner wall
of a pipe is not an efficient method to prevent acoustic energy propagation. The value of
the damping will increase for smaller pipe diameter, i.e., it can be of importance in
narrow pipes as found in catalytic converters (Allam (2004)).
On the influence of the internal sound attenuation on the measured reflection
coefficient
The experimental data presented in Figure 4 plotted as function of the Helmholtz number
ka for the case without flow and Strouhal number St = ka M for the flow cases) are
obtained using the cluster technique without iteration (two-microphone technique) and a
real wavenumber k=/c0 (no damping) and with iterations corresponding to the case for
which the measured damping is included in the experimental determination of the
reflection coefficient R. It can be seen from Figure 4 hat the Munts model predicts quite
well the trends of the measured data (typical error <10%).
18
1.1
Munt
Measured damping (Exp)
No damping (Exp)
1
0.9
|R|
0.8
0.7
0.6
No flow case
0.5
0.4
0.1
0.4
0.6
0.8
1
ka
1.2
1.4
1.6
1.8
(a)
1.2
Munt
Measured damping (Exp)
No damping (Exp)
1.1
1
|R|
0.9
0.8
0.7
M=0.07
0.6
0.5
1.5
10
15
St (ka/M)
20
25
(b)
Figure 4 (a), (b): Magnitude of the pressure reflection coefficient R at the opening of the
straight pipe measured with and without the effect of damping.
19
1.3
Munt
Measured damping (Exp)
No damping (Exp)
1.2
1.1
|R|
0.9
0.8
M= 0.18
0.7
0.6
4
St (ka/M)
Figure 4 (c) : Magnitude of the pressure reflection coefficient R at the opening of the straight
pipe measured with and without the effect of damping.
Figure 5 studies the effect of the choice of the flow speed (maximum, mean) for
computation of the reflection coefficient |R| using Munts model (Munt (1977, 1990)).
Measured damping
Munt (Mean Velocity)
Munt (Max Velocity)
1.4
1.2
|R|
1
0.8
0.6
M= 0.18
0.4
0.2
1500
2000
2500
3000
Frequency (Hz)
3500
20
Experimentally, the maximum velocity is obtained with a Pitot-tube aligned with the
centreline of the pipe. The relation between the maximum flow speed and the mean
velocity over the cross section is given in Schlichting (1968). At the Reynolds number
of this study, the mean velocity is 20% less than the maximum value. It can be seen
from Figure 5 that using the maximum speed seems to improve the predictions.
Munts model allows the determination of the complex value of the reflection
coefficient R at the opening. In this section, the phase of R (measured and predicted) is
converted into an end-correction term (see (2-18) and (2-19). Both experimental and
theoretical results are presented and shown in Figure 6. For the no flow case (Figure 6
(a)), prediction of the end correction agrees well with the experimental value for which
two types of excitation were tested: a broadband noise and steeped-sine excitation
ranging from 0 to 5 KHz. No discrepancies between the predicted and measured no flow
end correction appear even at low Helmholtz number for which experimental problems
can be encountered (room acoustic, low driving efficiency of loud-speaker, etc..). The
result for the low frequency limit obtained experimentally is very close to the value of
0.6133 derived by Levine & Schwinger (1948). An interesting feature is observed in the
high Strouhal number region. In this region ( St = ka M >> 1 ), one can see from
comparison with the no flow case that the flow does not change the value of the end
correction. It can also be seen that the measurements clearly support the drop in the end
correction for small St-number predicted by Munts model. For small St-numbers the
flow field plays an active part and coupling between the acoustic field and the flow field
is of importance. This coupling is imposed by Munt theoretically through the full Kutta
condition. This condition governs the transfer of energy between the acoustic field and
the unstable vortex sheet outside the opening.
21
Random excitation
Stepped-sine excitation
Munt
end correction
1
0.8
0.6
0.4
No flow case
0.2
0
0.23 0.4
0.6
0.8
1
1.2
ka
1.4
1.6
(a)
Experiment
Munt
1
end correction
0.8
0.6
0.4
M=0.1
0.2
0
1.5
3 4
8
10
St (ka/M)
12
14
16
(b)
Figure 6 (a), (b): End correction for a straight pipe.
22
1
Munt (Mean velocity)
Munt (Max velocity)
Experiment
0.9
0.8
end correction
0.7
0.6
0.5
0.4
0.3
0.2
M=0.15
0.1
0
1.1 2
7 8
St (ka/M)
10 11 12
(c)
Another investigation shown in Figure 6 (c) corresponds to the choice (mean or max) of
the value of the Mach number to be used in Munts model. The turnout is that the max
speed gives a better prediction up to St 4 5 and for higher St-number the mean speed
appears to be the best choice. The low frequency behaviour of the end correction has
been checked by using a larger microphone separation. The experimental normalised
end correction
St 0, M = 0.2 .
2.2.3 Sound absorption due to vorticity shedding
Measurement of the rate of the acoustic absorption related to vorticity production at the
pipe opening requires an estimation of the background noise due to the wall friction and
flow separation noise. Figure 7 shows the various levels of background noise. It is
necessary to ensure that the radiated power is of at least 10 dB greater than the
background noise generated by the flow and other rig noise. It is found from Figure 7
(a) that good SNR could not be obtained in the narrow region of 1.0 KHz. The
measurements presented below have been therefore restricted to Mach numbers less
than 0.2.
23
80
75
70
M=0.18
Loudspeaker+Flow noise
Lw (dB)
65
60
55
50
45
40
35
30
0
Figure 7: Various sound power levels externally measured with a rotating microphone.
The presence of mean flow in the pipe influences the rate of sound radiated to the far
field. The results presented in Figure 16 shows that acoustic losses due to vorticity is
dominantly a low frequency process. In contrast, sound attenuation through viscous
damping within the pipe is a high frequency mechanism. Bechert (1980) demonstrates
the connection between flow separation and sound reduction by conversion of the
acoustical energy into kinetic energy of the flow field. The upper frequency limitation
of the theory developed by Bechert (1980) is given as f < M c a and is indicated by
a vertical dashed line in Figure 8. The analytical model derived by Bechert (1980) is in
its form rather simplistic although good agreements are observed with our experimental
data. Linear effects are usually encountered in the near field where low frequency sound
absorption occurs. Thus, traditional and simple mathematical derivations are feasible in
this region. An analytical approach as the one described by Bechert (1980) can thus be
made.
24
20
10*log(Wrad/Wtrans)
10
0
-10
-20
-30
M=0.1
-40
-50
-60
Measurement
BECHERT
Theory validity
-70
0
1000
2000
3000
Frequency (Hz)
4000
5000
20
10log(Wrad/Wtrans)
10
0
-10
-20
-30
M=0.15
-40
-50
-60
-70
0
Theory validity
500
1500
2416
3500
Frequency (Hz)
Measurement
BECHERT
4500
Such a linear approach is not possible in the further downstream region where the flow
breaks into turbulence and sound energy is converted into heat. Turbulence creates
sound of quadrupole type much weaker than the absorption in the near field (dipole
sound sink). Another interesting mechanism pointed out by Bechert (1980) is the noise
conversion from narrow band to broader band in this region. The effects in this jet
mixing region are small and can be neglected.
25
2.3
Summary
This chapter has dealt with the sound propagation in an open straight pipe. Different
regions have been examined theoretically and experimentally and variation of the flow
speed has been performed up to a Mach number of 0.2. The internal damping through
viscous boundary layers has been measured with the means of a six-microphone array.
The measured results of the damping agree well with the Kirchoffs formula when no
flow is present in the pipe. However, for flow conditions, the formulation suggested by
Davies (Davies (1980)) based on Kirchoffs no flow formula underpredicts the internal
damping in the flow pipe. The experimental technique also allows to determine the
pressure reflection coefficient and impedance at the opening. The prediction is
conducted with a model derived by Munt (1977, 1990). A quite good agreement is
observed between the theoretical and experimental results even at relatively high flow
speeds.
The end correction derived from the experimental pressure reflection coefficient has
been compared with that of Munts model (Munt (1977, 1990)) derived for the
theoretical pressure reflection coefficient. The trends are well predicted and limits at
low frequency agree with the experimental data obtained at low frequency.
The acoustic losses caused by the shedding of vorticity at the opening of the flow pipe
has been experimentally obtained from induct and external measurements. The
analytical model of Bechert (1980) defines the opening as a distribution of monopole
and dipole sources. This model has been used to predict the sound absorption in the
vorticity shedding. Good agreement has been obtained with the experimental data.
26
3. Acoustics of tailpipes
3.1
The tailpipe experimental rig consists of several sections assembled as shown in Figure
9. The main difference compared to the rig used in the treatment of the aeroacoustics of
the straight pipe is the number of microphone used. Here, three microphone locations
are used in conjunction with the two microphone technique, whereas a six-microphone
array is used in the straight pipe rig in order to also measure the acoustic damping. The
arrangement is designed to permit flexibility between measurements. The microphones
are flush mounted to minimise flow disturbances. The cut-off frequency based on the
internal pipe radius (42 mm diameter, 2 mm thickness) for this pipe is 4.8 KHz. Pitottube results are shown in Table 1. A smooth inlet cone is used to minimise flow
separation noise at the inlet of the rig.
27
28
Figure 9: Layout of the experimental setup for tailpipes.
Elements
V1.max
V2, max
V3, max
Straight pipe
25
43
80
Oblique cut 0
25
43
80
Oblique cut 30
24
42
80
Oblique cut 45
24
42
80
Oblique cut 60
24
43
80
Diffuser 1
25
44
80
Diffuser 2.5
27
47
90
Diffuser 5
27
48
90
Diffuser 7.5
25
48
91
45 Bend L=0
23
41
80
45 Bend L=1D
24
41
80
45 Bend L=2D
24
42
80
45 Bend L=5D
24
42
80
90 Bend 1 R/D=1.6
23
41
78
90 Bend 2 R/D=2.5
24
41
78
Schematic diagrams and pictures of the various elements tested are now presented.
Three main categories are investigated: oblique cuts, diffusers and bends. Reflection
coefficients and sound power levels of the flow generated noise are compared with the
straight pipe case (reference).
Oblique cut
The steepness of the oblique cut at the outlet is given by the angle between the no cut
case (straight pipe) and the oblique cut. This angle has a value of 30, 45 and 60 as
shown in the following drawings (Figure 10). It should be noticed that the total length of
the element is constant (150 mm). The purpose of this cut is to reduce the reflection at
the outlet and thereby reducing the effect of resonances.
29
Diffusers
The second type of element investigated is the diffusing element (Figure 11). Slowly
divergent conical elements having a fixed length of 105 mm are studied. Measurements
on diffusers of semi angles of divergence of 1, 2.5, 5 and 7.5 are carried out. The
diffuser can reduce reflections at an opening and possibly also reduce flow noise
produced at the outlet.
30
45 bends
The bend elements consist of three parts, an upstream straight part, the bend and a
downstream straight pipe. The upstream distance between the inlet and the bend edges
31
has a fixed value of 70 mm (Figure 12). The downstream section located after the
curvature and extended to the outlet takes the length L of 0, 42, 82 and 210 mm. The
curvature radius for the bends is 50 mm. An angle of 45 is found between the upstream
and downstream sections.
L=0
L=D
L=2D
L=5D
L
90 bends
For the 90 bends the angle separating the upstream and downstream sections is 90
32
(Figure 13). Two radii of curvature (67 mm and 105 mm) at the bends permit ratios of
curvature radius R to pipe diameter D to be of 1.6 and 2.5. The interest here is on the
aeroacoustics of bends close to the tailpipe opening.
R/D=1.6
R/D=2.5
R
D
Figure 13: 90 bends.
Theoretical investigation of the reflection coefficient and impedance at the inlet and
outlet of the elements differs slightly between the types of elements. Each case is now
theoretically detailed.
The main approach is to calculate the reflection coefficient at the outlet of the element
using Munts model (Munt (1977, 1990)) and to move the results to the inlet through its
transfer matrix calculated from SID (Nygrd (2000)) (Sound in Duct software
developed at MWL).
The outlet impedance Zout is calculated after derivation of the outlet reflection
coefficient Rout using the relation
1 + Rout
Z out =
1 Rout
0 c0
,
S
(3-1)
where Rout is calculated from Munts model (Munt (1977, 1990)) based on the radius
and mean Mach number M at the outlet for diffusers. For the oblique cuts, it was found
preferable to use the inlet element radius.
33
The transfer matrix relating the pressure p and volume flow speed q between the inlet
and outlet of the element is of the form
pin T11 T12 pout
.
=
(3-2)
(3-3)
Using (3-1), the reflection coefficient Rin at the inlet can similarly be written
Rin =
Z in S ( 0 c0 ) 1
.
Z in S ( 0 c0 ) + 1
(3-4)
The elements T11, T12, T21 and T22 are obtained from the transfer matrix calculated using
SID.
Regarding the bends, no transfer matrix calculation is done; the reflection coefficient is
shifted to the outlet through a straight pipe with same length.
Oblique cut case
It was primarily assumed that only the increase in the cross-section area at the outlet
dominated the behaviour of the reflection coefficient. Therefore an equivalent radius has
been calculated for this opening and used with Munts model for same Mach number.
However, the reflection coefficient obtained from Munts model with this calculated
radius did not agree well with the experimental data at high frequency. This discrepancy
can be explained by the additional acoustical effect of the obliquity. As opposed to the
other elements, the acoustical wave does not encounter the lips of the oblique cut at the
same time, thus the wave is reflected at different time along the oblique cut introducing
a phase difference. To model this, it was decided to split the oblique cut in a number of
parallel pipes with varying length as shown in Figure 14.
Figure 14: Pipe network for oblique cut transfer matrix calculation.
34
The theoretical transfer matrix for diffusers is derived from the software SID (Sound in
Duct). The methodology used in SID is to split the overall diffuser into straight pipe
segments connected to each other in serial. The user inputs for any diffuser elements are
the inlet and outlet cross-section, the length and the number of straight pipe segments.
Knowing the transfer matrix for each straight pipe and applying continuity of pressure
and volume flow at the interface between two consecutive pipe elements, the transfer
matrix for the entire network is derived. The centreline distance inlet-outlet, i.e., the
length, for the diffusers has a fixed value of 105 mm. A number of 10 segments is
chosen for the splitting as shown in Figure 15.
By combining the outlet pressure reflection coefficient Rout predicted from Munts
model with the diffuser transfer matrix, we can derive the inlet pressure reflection
coefficient Rin and impedance Zin.
Bend case
No transfer matrix calculation is performed for the bend cases. The reflection
coefficient is calculated at the outlet of the bend using Munts model and moved to the
plane of interest by a centreline distance separating these two points. In other word, the
35
bend curvature is not taken into account for the prediction of the internal noise
propagation in bends.
3.1.4 Experimental procedures
The experimental technique to measure the pressure reflection coefficient R for all
elements is based essentially on the two microphone technique (bom & Bodn
(1988)). The separations between the microphones in the rig yield a frequency range
(see (1-2)) as follows
Microphone 1 and
Microphone 1
2, s12=28 mm
and 3, s13=170
mm
Two- microphone
frequency criteria (Hz)
3.2
610 f 4900
100 f 800
Acoustic properties of the tailpipe geometries and the deviation from the straight pipe
case are now studied. The circular straight pipe termination is replaced by various
elements; oblique cuts, diffusers and bends as described in the previous section.
Influence of the termination geometry and flow speed on the reflection coefficient and
impedance and flow noise generated is investigated with comparison to the reference
case, the straight pipe.
3.2.1 Reflection coefficient and impedance
Results for the reflection coefficient and impedance at the outlet of the tailpipe elements
described in this section are now presented. The impedance for the tailpipe elements are
moved assuming a straight pipe from the reference position (microphone 1) to the outlet
opening. This procedure, which gives an equivalent radiation impedance, enables a
direct comparison with the reference straight pipe case. Experimentally, it is obtained
using three microphones with the one located the nearest to the pipe open end chosen as
reference. Data are presented up to the straight pipe cut-off frequency (4800 Hz). Thus,
only plane waves are considered in this analysis. An overview of experimental data
collected for the elements are presented, and the influence of mean flow speed M is
studied.
36
Oblique cuts
Influence of the obliquity angle without flow
Figure 16 shows the experimental reflection coefficients for the oblique cuts for various
angles and for no flow condition. At low frequencies (<1500 Hz), all the elements
behave as a straight pipe. It can be explained by the relatively large wavelength
compared to the opening characteristic dimension. In other words, at low frequency, the
sound propagation is not influenced by the geometry of the elements. Bechert (1980),
Peters and al. (1993) report the same observation.
Deviation between the elements is observed at high frequencies. The decrease in the
reflection coefficient as the obliquity angle increases is related to the successive
removal of a distinct cross section where the pipe terminates.
1.5
Straight pipe
30
45
60
|R|
0.5
37
Straight pipe
30
45
60
No flow case
Real(Z)
0
500 1000 1500 2000 2500 3000 3500 4000
Frequency(Hz)
(a)
2
No flow case
Imag(Z)
0
-1
Straight pipe
30
45
60
-2
-3
-4
(b)
Figure 17: No flow opening impedance experimental data for oblique cuts.
38
As for the reflection coefficient, both the real and imaginary parts of the impedance tend
to zero in the low frequency limit as for the straight pipe. It appears that the real part is
not affected by the change of the obliquity. A deviation can be noticed, however, for the
mid-frequency range when considering the imaginary part. A peak is observed for the
sharpest cut (60) for which a resonance seems to occur.
Effect of mean flow
Figure 18 shows the experimental values for the reflection coefficients and impedances
for the oblique cut elements of different angles and straight pipe for flow Mach number
of 0.19. Figure 18 shows that the general behaviour of the oblique cut elements is not
ruled by the presence of the flow in the pipe.
1.4
Straight pipe
30
45
60
1.2
1
|R|
0.8
M=0.19
0.6
0.4
0.2
0
(a)
Figure 18 (a): Experimental reflection coefficient for flow oblique cuts.
39
Straight pipe
30
45
60
Real(Z)
6
5
M=0.19
4
3
2
1
0
-1
100
1000
2000
3000
Frequency (Hz)
4000
(b)
3
2
M=0.19
1
Imag(Z)
0
-1
-2
Straight pipe
30
45
60
-3
-4
-5
-6
100
1000
2000
3000
Frequency (Hz)
4000
(c)
Figure 18 (b), (c): Experimental impedance for flow oblique cuts.
40
Theoretical investigation
1.5
=30
|R|
0.5
M=0
0
150
1000
2000
Frequency (Hz)
3000
4000
The prediction by the transfer matrix model including the geometrical effects agrees
reasonably well with the experimental data. At even higher frequencies, the transfer
matrix theory breaks up and deviates from the experimental result. A more refined
model is certainly needed. shows the results obtained for the model including Munts
theory (Munt (1977, 1990)) and transfer matrix calculation.
1.5
=45
Munt+SID
Experiment
|R|
0.5
No flow case
0
0
1000
2000
3000
Frequency (Hz)
4000
41
0.1
Experiment
Munt+SID
0
=45
-0.1
-0.2
No flow case
Imag(Z)
-0.3
-0.4
-0.5
-0.6
-0.7
-0.8
-0.9
150
1.2
1
Real(Z)
0.8
1000
2000
3000
Frequency (Hz)
4000
Experiment
Munt+SID
=45
0.6
0.4
0.2
0
-0.2
No flow case
-0.4
150
1000
2000
3000
Frequency (Hz)
4000
Figure 21: Modelling the opening reflection coefficient and impedance of oblique cuts.
42
Diffusers
The diffuser elements are now studied. Experimental data are presented for the diffusers
and the straight pipe. It is followed by a comparison using the transfer matrix
methodology. Discrepancies between the experiment and measured data are explained
by a series of measurements performed at low frequencies.
Influence of diffuser half-angle
Figure 22 (no flow), Figure 23 (flow case) present the experimental results obtained for
1.5
No flow case
5
2.5
1
Straight pipe
|R|
0.5
0
100
1000
2000
Frequency (Hz)
3000
Figure 22: Diffuser experimental results at outlet (reflection coefficient) and no flow.
43
1.2
No flow case
Real(Z)
0.8
0.6
Straight pipe
1
2.5
5
0.4
0.2
0
-0.2
100
1000
2000
3000
Frequency (Hz)
0.8
0.7
No flow case
Imag(Z)
0.6
0.5
0.4
0.3
Straight pipe
1
2.5
5
0.2
0.1
0
-0.1
100
1000
2000
3000
Frequency (Hz)
44
1
0.8
M=0.09
0.6
Real(Z)
0.4
0.2
0
-0.2
Straight pipe
1
2.5
5
7.5
-0.4
-0.6
-0.8
-1
150 500
1000
1500
2000 2500
Frequency (Hz)
3000
3500
1
0.8
M=0.09
0.6
Imag(Z)
0.4
0.2
0
-0.2
5
1
Straight pipe
2.5
7.5
-0.4
-0.6
-0.8
-1
150 500
1000
1500
2000
2500
Frequency (Hz)
3000
3500
Figure 23: Diffuser experimental results at outlet (impedance) and with flow.
45
Straight pipe
1
2.5
5
7.5
1.4
M=0.09
1.2
|R|
1
0.8
0.6
0.4
0.2
150
500
1000
1500
2000
2500
Frequency(Hz)
3000
3500
Figure 23: Diffuser experimental results at outlet (reflection coefficient) and with flow.
One interesting feature observed from the flow case is the significant increase for low
frequencies of the reflection coefficient. Instead of approaching 1 as it always does for
the straight pipe, it seems to approach a value larger than 1.
46
Flow effects are shown for a specific diffuser angle in Figure 24 (1) and Figure 25(7.5).
1.6
M=0
M=0.09
M=0.19
1.4
1.2
=1
real(Z)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
500
1000
1500
2000
2500
Frequency (Hz)
3000
1.5
3500
4000
M=0
M=0.09
M=0.19
=1
Imag(Z)
0.5
-0.5
-1
500
1000
1500
2000
2500
Frequency (Hz)
3000
3500
4000
Figure 24: Influence of the Mach number on the outlet impedance of the diffuser 1.
47
1.6
1.4
=7.5
1.2
Real(Z)
1
0.8
0.6
0.4
0.2
0
M=0
M=0.09
M=0.19
-0.2
-0.4
100
500
1000
1500
2000
2500
Frequency (Hz)
3000
3500
4000
(a)
1.5
=7.5
Imag(Z)
0.5
-0.5
M=0
M=0.09
M=0.19
-1
100
500
1000
1500
2000
2500
3000
3500
4000
(b)
Figure 25(a), (b): Influence of the Mach number on the outlet impedance of the diffuser 7.5.
48
1.8
M=0
M=0.09
M=0.19
=7.5
1.6
1.4
|R|
1.2
1
0.8
0.6
0.4
0.2
500
1000
1500
2000
2500
Frequency (Hz)
3000
3500
4000
(c)
Figure 25 (c): Influence of the Mach number on the pressure reflection coefficient and outlet
impedance of the diffuser 7.5.
It can be seen that an increase of the Mach number results in a reduction of the radiation
resistance (real part of the load impedance Z) at low frequency. The end correction
(imaginary part of the load impedance Z) is on the other hand unchanged at low
frequency. At high frequency, the resistive term is not dependent on the flow speed
whereas the end correction increases with the flow Mach number. The two cases
presented here (also at low frequency, see Figure 26 and Figure 27) show that both the
angle and flow influence the acoustics of the termination.
Diffuser low frequency measurement
49
1.5
Diffusers: 1, 2.5, 5, 7.5
Straight pipe
|R|
0.5
M=0
0
20
50
100
150
Frequency (Hz)
(a)
200
1.5
(1+M)/(1-M)
1.13
1
|R|
Straight pipe
0.5
M=0.06
0
20
50
100
150
Frequency (Hz)
(b)
200
Figure 26 (a), (b): Low frequency reflection coefficient measurements for diffusers (M=0, 0.06).
50
1.5
(1+M)/(1-M)
1.27
1
Diffusers:
|R|
Kutta limit
Straight pipe
0.5
M=0.12
20
50
100
150
Frequency (Hz)
200
1.6 (1+M)/(1-M)
1.47
Diffusers:
Kutta limit
Straight pipe
M=0.19
0
20
50
100
150
Frequency (Hz)
200
Figure 26 (c), (d): Low frequency reflection coefficient measurements for diffusers (M=0.12,
0.19).
51
Hirschberg and al. (1988) have studied the vortex sound generated by diffusers. They
use the energy reflection coefficient RE defined as
2
RE = R B = R
(1 M ) ,
(1 + M )
2
(3-5)
where RB is the enthalpy reflection coefficient and R is the sound pressure reflection
coefficient used throughout this paper. It indicates that for RE exceeding unity, sound is
amplified by the flow separation mechanism. As seen from Figure 26 the tested diffusers
are not producing extra sound. But for angles around 5-7.5 we approach the limit
( (1 + M ) (1 M ) of no losses, i.e., all the incident acoustic power is reflected for low
frequencies.
Diffuser impedance results
The real and imaginary part of the impedance is plotted in Figure 27 as function of the
Helmholtz number ka , the diffuser angle and flow Mach number M. The data was
obtained at the microphone reference 1 and moved to the open end by a distance of
l=150 mm (cf Figure 1) corresponding to the opening of the straight pipe and to the inlet
of the diffusing elements. The length of the elements is constant (105 mm).
For results with no flow (M=0), no deviations are observed between the diffusers. It
shows that the diffusers have the same behaviour when no airflow is introduced in the
pipe.
In view of the flow results, M=0.05 and 0.2, the trend is for the real part of the
impedance (radiation resistance) to decrease as the diffuser angle and Mach number
M increase. This observation is of paramount importance when the diffuser is connected
to an exhaust line. An optimal design of exhaust system requires to minimise such
radiation. Modelling and performance estimation of exhaust line with these measured
load impedance data are presented in the section Modelling of exhaust lines.
52
0.06
0.05
Real(Z)
0.04
0.03
M=0
Diffusers
0.02
0.01
0
-0.01
0.01
Straight pipe
0.02
0.03
0.04
ka
0.05
0.06
0.07
0.06
0.07
0.5
Imag(Z)
0.3
Diffusers
0.1
0
-0.1
-0.3
-0.5
0.01
Straight pipe
M=0
0.02
0.03
0.04
ka
0.05
(a)
Figure 27 (a): Normalised inlet impedance measurement at low Helmholtz number for diffuser
elements (M=0).
53
M=0.05
Straight pipe
Real(Z)
-0.02
diffusers:
-0.04
-0.06
-0.08
0.01
0.02
0.03
0.04
ka
0.05
0.06
0.07
0.4
0.3
M=0.05
Imag(Z)
diffusers:
0.2
0.1
Straight pipe
0
-0.05
0
0.01
0.02
0.03
ka
0.04
0.05
0.06
0.07
(b)
Figure 27 (b): Normalised inlet impedance measurement at low Helmholtz number for diffuser
elements (M=0.05).
54
0.1
M=0.2
Straight pipe
Real(Z)
0
-0.05
diffusers:
-0.15
-0.25
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.07
0.08
ka
0.45
M=0.2
Imag(Z)
0.35
0.25
0.15
diffusers:
0.05
-0.05
0.01
Straight pipe
0.02
0.03
0.04
0.05
0.06
ka
(c)
Figure 27 (c): Normalised inlet impedance measurement at low Helmholtz number for diffuser
elements (M=0.2).
55
1.5
Experiment
Munt +SID
=5
|R|
0.5
No flow case
1.5
Munt+SID
Experiment
=7.5
|R|
0.5
No flow case
0
56
1.5
Exp
Munt+SID
=2.5
|R|
0.5
M=0.05
0
=2.5
Munt+SID
Experiment
|R|
0.5
M=0.19
0
500
1.4
Experiment
SID
1.2
|R|
1
=7.5
0.8
0.6
0.4
M=0.09
0.2
0
500
Figure 29: Theoretical and experimental comparison for reflection coefficient for flow diffusers.
57
The results presented in Figure 28 from the transfer matrix model used to predict the
reflection coefficient for the no flow case agrees well with the experimental data.
Modelling of the standing wave pattern (undulation) is achieved theoretically. There is a
significant discrepancy between the calculated and experimental values for lower
frequency when flow is present in the rig (Figure 29). Dequand and al. (2002) showed
for diffusers of this type that the flow separation point is not fixed but moves along the
inner walls of the diffuser elements. Munts model to calculate the pressure reflection
coefficient at the outlet cross-section is based on a sharp-edged condition (Kutta
condition). Flow separation is assumed to occur solely at the lips in Munts model. Due
to the flow separation within the diffuser element for half-angle greater than 4, the
boundary layer at the opening is substantially thicker violating the simple flow
separation model used by Munt.
Bends
58
45 bends
1.5
Straight pipe
L=0
L=1D
L=2D
L=5D
45 bends
|R|
0.5
No flow case
0
100
1000
2000
3000
Frequency (Hz)
4000
1.5
45 bends
|R|
0.5
Straight pipe
L=0
L=1D
L=2D
L=5D
M=0.19
0
200
1000
2000
3000
Frequency (Hz)
4000
(a)
Figure 30 (a): Reflection coefficient measurements for 45 bends.
59
45 bends
Straight pipe
L=0
L=1D
L=2D
L=5D
2.5
2
Imag(Z)
1.5
1
M=0.19
0.5
0
-0.5
100
1000
2000
3000
Frequency (Hz)
45 bends
Straight pipe
L=0
L=1D
L=2D
L=5D
2.5
2
Real(Z)
4000
1.5
1
M=0.19
0.5
0
-0.5
100
1000
2000
3000
Frequency (Hz)
60
4000
90 bends
1.5
Straight pipe
R/D=1.6
R/D=2.5
Munt
90 bends
1
|R|
0.5
No flow
0
1.5
R/D=1.6
Straight pipe
R/D=2.5
Munt
90 bends
1
|R|
0.5
M=0.06
0
61
Straight pipe
R/D=1.6
R/D=2.5
No flow case
Real(Z)
0.5
0
-0.5
-1
500 1000 1500 2000 2500 3000 3500 4000
Frequency (Hz)
Imag(Z)
No flow case
R/D=1.6
Straight pipe
R/D=2.5
0
-1
-2
-3
500 1000 1500 2000 2500 3000 3500
Frequency (Hz)
(b)
Figure 31 (b): Impedance measurements for 90 bends (M=0).
62
1.5
Straight pipe
R/D=1.6
R/D=2.5
M=0.19
Real(Z)
0.5
0
-0.5
-1
-1.5
500
1.5
M=0.19
1
Imag(Z)
0.5
0
-0.5
Straight pipe
R/D=1.6
R/D=2.5
-1
-1.5
500
63
3.3
Summary
64
65
The volume flow speed q2 at the outlet of the exhaust line can be calculated by defining
(Bodn (1989)) the source impedance Zs, the source pressure p S (source pressure), the
transfer matrices of the elements positioned between the source and the termination and
the load impedance. Comparison of the volume flow speed q2 from a reference system
enables the determination of the Insertion Loss (IL).
The pressure p and volume flow speed q at the inlet (1) and outlet (2) of the exhaust
line depicted in Figure 32 is given by
p1 = p S Z S q1 ,
p2 = Z L q2 .
(4-1)
The state variables pressure p and volume flow speed q at the inlet (1) and outlet (2)
of the entire system is related for any exhaust line through its transfer matrix. The
T12
T
depends on the elements composing the
exhaust line transfer matrix Texh = 11
T21 T22
line between the source and the termination. The relation between the pressure p and
volume flow speed q at the inlet (1) and outlet (2) is given as
p1 T11 T12 p2
=
.
q1 T21 T22 q2
(4-2)
(4-3)
The complex pressure reflection coefficient Rref is measured at the in-duct flushmounted reference microphone using the two-microphone technique. The value of the
reflection coefficient RL at the inlet is calculated using (1-3).
The load impedance used in the calculation of the IL for the exhaust lines is formulated
as
1 + R L 0 c0
,
Z L =
1 RL S
(4-4)
where S is the inlet cross-section area. The results for the normalised load impedances
are given in Figure 27.
Due to the low frequency investigation, it is assumed throughout this section that the
66
sound radiated to the far field is of monopole type, i.e., it is proportional to the volume
flow speed q at the outlet of the entire system. The low frequency range studied also
implies that the inlet tailpipe impedance is equivalent to the outlet impedance.
Combining (4-1) and (4-3) gives the monopole source strength q2 at the outlet (2) of
the exhaust line
q2 =
4.2
p S
.
Z LT11 + T12 + Z S Z LT21 + Z S T22
(4-5)
For the case studied below a Scania truck exhaust line with L1=1.3m and L2=2.6m is
used (Glav (1994)).
The total transfer matrix for the entire system in Figure 33 is calculated from
T
Texh = Tsp .L1 Tsilencer Tsp ,L2 = 11exh
T21exh
T12 exh
,
T22 exh
(4-6)
where the straight pipe transfer matrix Tsp is given in equation (4-11).
The frequency independent source impedance Z S = 0 c0 exp(i 7 / 4 ) S where S is the
pipe cross-section area is taken from Callow & Peat (1988). No source strength Ps is
required and can be set up to any value as it cancels out when deriving the expression
for IL.
67
Tsilencer = 11 sil
T21sil T22 sil
(4-7)
provided by the Swedish heavy-vehicle manufacturer Scania for no flow condition gives
the following transmission loss
1
T
TL = 10 log10 T11 sil + 12 sil + Z 0T21sil + T22 sil
4
Z0
(4-8)
The inlet and outlet impedances are equal to Z 0 = 0 c0 S with S taken as the inlet and
outlet silencer cross-section area and equal to 8.66 10 3 m2. A plot of the Scania
silencer transmission loss is given in Figure 34.
60
Measured TL (dB)
50
40
30
20
10
0
3050
100
350
400
The transfer matrices for the two pipes Tsp ,L1 and Tsp ,L2 in Figure 33 are computed using
(4-11) with L1=1.3m and L2=2.6m.
In order to derive the IL, it is necessary to define a reference system. It is defined as
shown in Figure 35 as a source, a pipe of zero length with transfer unit matrix and the
measured termination impedance ZL
68
T11ref
Tref =
T21ref
T12 ref 1 0
=
.
T22 ref 0 1
(4-9)
Insertion loss
An insertion loss index IL is calculated for the system in Figure 33 and the reference
system in Figure 35 with the same load impedance ZL. The IL is defined as the ratio of
the source strength calculated for the automotive exhaust line in Figure 33 ( qexh ) to the
reference line in Figure 35 ( qref ).
IL
exh
= 20 log
q
exh
10 q
ref
20
log
10 Z
T
+T
+Z Z T
+Z T
11ref
12 ref
S L 21ref
S 22 ref
T
+T
+Z Z T
+Z T
11exh
12 exh
S L 21exh
S 22 exh
(4-10)
where T11ref, T12ref, T21ref, T22ref elements are given in (4-9), T11exh, T12exh, T21exh, T22exh
elements are obtained from (4-6) and ZL is the measured tailpipe load impedance.
The measured load impedance ZL is obtained on elements with an inlet diameter of 42
mm. The silencer data provided by Scania corresponds to an inlet and outlet diameter of
105 mm. It was thus necessary to obtain the load impedance for the diffuser using the
same dimensionless Helmholtz number ka based on either the silencer or pipe radius.
Results
The IL calculated from (4-10) for systems terminated with the diffuser elements of
different semi-angle are shown in Figure 36 for the no flow condition. It can be seen
from Figure 36 that changing the diffuser element at the termination of the automotive
exhaust line do not significantly modify the IL results in the case of no airflow. The
large dip ( 55 Hz ) comes from the first half-wavelength resonance in the tailpipe L2.
However, when flow is introduced, it can be seen from Figure 37 that the larger the
diffuser half angle , the larger the IL value at the resonance frequency. An increase in
69
35
=1
=2.5
=5
=7.5
M=0
30
IL (dB)
25
20
15
10
5
20
30
40
50
60
70
Frequency (Hz)
80
90
100
Figure 36: No flow IL for the automotive exhaust line terminated with diffusers.
35
M=0.2
30
IL (dB)
25
20
=1
=2.5
=5
=7.5
15
10
20
30
40
50
60
70
80
90
100
Frequency (Hz)
Figure 37: IL with flow for the automotive exhaust line terminated with diffusers.
70
Looking at the largest angle (=7.5) investigated, Figure 38 shows the improvement at
different Mach numbers.
35
M=0.06
M=0.1
M=0.2
diffuser 7.5
30
IL (dB)
25
20
15
10
5
20
30
40
50
60
70
Frequency (Hz)
80
90
100
Figure 38: Automotive exhaust line IL for the diffuser 7.5 at various flow Mach number M.
It appears that the IL is greater when flow Mach number and diffuser angle increase.
In order to understand these results on the tailpipe resonance, a simplified system
consisting only of a tailpipe is studied next. Since the tailpipe at the upstream end is
normally connected to a large volume (silencer), it will be assumed that the source
impedance ZS is zero.
4.3
This system comprises a pipe of length L and a source strength Zs=0 as shown in Figure
39.
71
A reference system is taken for the single pipe exhaust line terminated with diffuser of
angle =1 to ensure overall fixed length. The diffusers have a constant length of 105
mm.
The transfer matrix for a straight pipe element with flow of length L is
Tsp ,L
1
[exp (ik + L ) + exp ( ik L )] 0 c0 [exp (ik + L ) exp ( ik L )]
2S
,
= S2
1
0 0
(4-11)
An insertion loss IL is calculated and is defined as the ratio of the source strength
calculated for the exhaust line in Figure 39 ( qsp ) terminated with diffusers to the same
exhaust line but with the diffuser of angle 1 ( qref ).
q sp
ILsp = 20 log
10 q ref
Z
+T
+Z Z
+Z T
T
T
S L ref 21sp , L
S 22 sp , L
12sp , L
= 20 log L ref 11sp , L
10 Z
+T
+Z Z
+Z T
T
T
S Ltailp 21sp , L
S 22 sp , L
12 sp , L
L tailp 11sp , L
(4-12)
The elements T11sp,L, T12sp,L, T21sp,L, T22sp,L are obtained theoretically from (4-11). The
measured load impedance for the 1 diffuser (ZLref , reference case) and tailpipe ZLtailp are
taken from the measurement of the pressure reflection coefficient RL and (4-4).
Results
Figure 40 shows the effect of the diffuser angle on the IL for the single pipe exhaust
line with flow Mach number M=0.2.
An improvement at the tailpipe resonance frequencies nc0 2 L , n = 1,2 is observed. The
results are consistent with the observations made on the automotive exhaust line. It is
immediately apparent from Figure 40 that a diffuser with an angle larger than 1
improves the IL at the tailpipe resonance frequencies of the pipe length L.
Figure 41 displays the influence of the flow Mach number on the IL for the diffuser
angle =7.5.Figure 40 and Figure 41 illustrate and confirm the results obtained from the
automotive exhaust line, i.e., the IL at the tailpipe resonance frequencies increase with
the diffuser angle and flow Mach number M.
72
M=0.2
=2.5
=5
=7.5
IL (dB)
(reference : diffuser 1)
-2
20
40
60
80
100
120
140
Frequency (Hz)
Figure 40: Influence of the diffuser angle for the single pipe exhaust line.
7
6
5
M=0.06
M=0.1
M=0.2
IL (dB)
4
3
2
1
0
-1
-2
-3
20
60
80
100
Frequency (Hz)
120
140
Figure 41: Single pipe exhaust line IL for the diffuser 7.5 at various Mach number M.
73
4.4
Summary
The investigation of the acoustics of diffusers as pipe terminations has showed low
frequency deviations from the straight pipe case. The magnitude of the pressure
reflection R at the opening of the diffusers has been experimentally obtained using the
two-microphone technique. It has been found that at low frequency (< 200 Hz) and with
airflow in the rig, the magnitude of the pressure reflection coefficient R is greater for the
diffusers than for the straight pipe.
The experimental load impedance results for diffusers have been implemented in the
calculation of the insertion loss IL for two types of designs; a Scania truck exhaust and a
single pipe system.
In view of the results obtained for both exhaust lines terminated with diffuser elements,
it can be concluded that the magnitude of the IL is increased at the tailpipe resonance
frequencies. The effect increases with the diffuser angle and flow Mach number and
improvements of up to 10 dB seems possible in practice.
Another important parameter to consider when designing such system is the sound
generation due to flow separation. This is treated in the next chapter.
74
The comparative method (ISO 3747) was chosen for determination of noise generated at
the opening of the rig due to the flow separation. This measurement technique uses a
reference sound source in a reverberant room for which the sound power level is known.
The reference sound source power levels are provided in 1/3-octave bands. A half-inch
microphone mounted on a rotating support measured and averaged sound pressure
levels of the signal either generated at the opening or from the reference source into the
diffuse field.
75
Flow noise is created in different regions of a straight pipe (Figure 42). The following
section aims to compare the sound power levels of the noise produced inside the pipe
(1), at the lips (2) and in the jet mixing region located outside the pipe (3).
Starting with the internal flow noise in a straight pipe, a semi-empirical model (VDI
3733 (1996)) is presented and used to determine the internal sound power level
generated by the flow inside the pipe.
The VDI 3733 guideline gives a semi-empirical model obtained from flow noise induct
measurements on long pipes with anechoic terminations. For Mach number (M=U/c0)
of less than 0.3, the internal noise generation due to the flow separation inside a straight
pipe is dominantly of dipole type (VDI 3733) implying that the flow induced sound
power is proportional to U6. The flow generated sound power level inside a straight pipe
is empirically formulated as
W
LWi = 10 log 10
W0
U
= K + 10 log 10
U0
S
+ 10 log10
S0
p
NT
+ 10 log10 25 log10
p0
N 0 T0
15 log 10
(5-1)
where K=8-0.16U (correction factor) and the reference values are W0=10-12 W, U0=1
m.s-1 (flow speed), S0= 1 m2 (cross-section area), p0= 101325 Pa (pressure), T0=273 K
(temperature), N0= 287 J.kg-1.K-1 (specific gas capacity), 0=1.4 (specific heat ratio).
The internal flow induced sound power level Lwi,1/3 in a straight pipe in 1/3-octave bands
is expressed as (VDI 3733).
LWi , 1 / 3 = LWi + LW , oct 4.77 dB ,
76
(5-2)
(5-3)
(5-4)
The following empirical formulation (Davies and al. (1980)) is used to predict the
pressure reflection coefficient R.
(5-5)
(5-6)
),
(5-7)
where LWi ,1 / 3 is given by (5-2) and corresponds to the empirically predicted internal
sound power in 1/3-octave bands.
Figure 43 shows the flow noise measured in the far field together with the internal noise
results obtained from (5-7). The flow noise measurement corresponds to the flow noise
generated from the inside, from the lips and jet mixing region and picked up by the
rotating microphone. The internal flow noise sound power level calculated from (5-7)
(VDI 3733) indicates that the flow noise created inside the pipe and transmitted out is
not the dominant mechanism for our external measurement. It appears that the measured
data correspond to the noise created at the lips and in the jet mixing region under the
Mach number range investigated.
77
120
M=0.06 (Exp)
M=0.1 (Exp)
M=0.19 (Exp)
M=0.06 (VDI 3733)
M=0.1 (VDI 3733)
M=0.19 (VDI 3733)
100
Lw (dB)
80
60
40
20
Figure 43: Comparison between the internal noise produced inside the pipe calculated from VDI
3733 with the measured flow noise in the reverberant room.
Kuhn & Morfey (1976) argue that the jet mixing noise produced outside the pipe
becomes dominating compared to the noise generated within the pipe and at the lips for
flow Mach numbers varying from 0.18 to 0.56. For the tests carried out in this work
with Mach numbers up to 0.2, it seems probable that the jet mixing noise (quadrupole)
is weaker than the dipole type of noise generated at the lips.
5.1.3 Scaling laws
(5-8)
where W is the flow induced noise sound power, U is the flow speed, D is the pipe
diameter, is the coefficient determining the aeroacoustic source type, F(St) is the
dimensionless spectra to be calculated for each element, St = f U D is the Strouhal
78
M are plotted.
A routine is implemented to obtain the optimal value of for which minimum standard
deviation between the plots is achieved. This procedure is required to obtain only one
scaling law for each element.
The value of the coefficient in (5-8) depends on the type of frequency bands used
(1/3-octave band analysis has been performed in this paper), the type of source
(monopole, dipole, quadrupole) and whether the flow noise generated is internally or
externally.
It is usually suitable to present the data in logarithmic scale as
Ls = 10 log10 ( F ( St )).
(5-9)
Others authors prefer the use of the formulation derived by Nelson & Morfey (1981)
having the following form
c3S
Ls ,f ,NelsonMorfey (St ) = Lw ,f 10 log10 0 0
16W
ref
10 log10 (C L2 ) 10 log10 (M 4 ).
(5-10)
The dimensionless spectrum Ls,f given in equation (5-10) is only applicable in the
plane wave range, that is, below the pipe cut-off frequency fc, which for a pipe of
diameter D is
fc =
1.84co
.
D
(5-11)
4.8 KHz). Comparisons with other authors scaling laws require the conversion of the
Nelson and Morfeys formula (5-10) to the form (5-8) used throughout this analysis.
Using (5-8), (5-9) and (5-10) the conversion (for =1 in (5-8) ) is found to be
Ls = Ls ,NelsonMorfey 13 dB .
(5-12)
Nygrd (2000) applied a correction of 3 dB due to the effect of the walls on the in-duct
sound pressure level measurement he performed. A quasi-reverberant field is caused by
the presence of the pipe walls resulting in an excess level of the sound pressure level by
a factor of 0 to 6 dB. VDI 3733 suggests setting this correction factor by 3 dB.
Determination of the pressure loss coefficient CL
Table 2 gives the pressure loss coefficient CL calculated from the measured pressure
Inlet mean
Velocity U (m/s)
Pressure loss
coefficient CL
Diffuser 1
25 44
80
6 104
9.9 104
18 104
0.25
Diffuser 2.5
27 47
90
6.1 104
11 104
20 104
0..48
Diffuser 5
27 48
90
6.1 104
11 104
20 104
0.37
Diffuser 7.5
25 48
91
5.6 104
9.9 104
20 104
0.55
24 41
80
5.4 104
9.2 104
18 104
0.17
24 42
80
6 104
9.2 104
18 104
0.16
24 42
80
6 104
9.2 104
18 104
0.23
23 41
78
6 104
9.2 104
18 104
0.25
24 41
78
6 104
9.2 104
18 104
0.24
45 Bend
L=1D
45 Bend
L=2D
45 Bend
L=5D
90 Bend 1
R/D=1.6
90 Bend 2
R/D=2.5
80
The pressure loss coefficient CL indicates the pressure difference between the inlet and
the outlet of an element. For elements (flow constrictions) inside a pipe producing flow
noise, CL should be used in the scaling law as proposed by Nelson & Morfey (1981).
However, for tailpipe terminations, it is not clear that this is useful and therefore CL can
be omitted in the scaling laws. The flow noise investigation requires the measurement of
the airflow pressure drop P between the inlet of the elements and the room (Figure 9).
It was carried out using an averaging method, where four circumferential holes were
drilled in the pipe to measure simultaneously the static pressure difference between the
inlet and the outlet of the elements. In turbulent flow the pressure loss coefficient
C L = 2 P 0U 2 is essentially independent of the flow speed U and its magnitude
depends solely on the geometry of the elements (Blevins (1984)).
Scaling laws for bend cases
The flow generation process for diffusers occurs at the outlet of the elements whereas
the sound source of interest to be scaled for bends is located internally at the bends. The
sound power level measured is performed externally and corresponds to the flow
induced at the bend and at the opening. Thus, two corrections (open end flow noise
removal, end reflection) are necessary to compute the proper sound power level
generated by flow separation in bends. To remove background noise, i.e., noise
produced along the pipe and at the opening, the measured flow induced sound power
level of the straight pipe is subtracted from the sound power level measured for each
bend. However, part of the flow induced sound generated inside the element is reflected
back at the opening as shown in Figure 44. This is corrected using equation (5-7).
81
5.2
Measurement of flow noise in 1/3-octave band in the reverberant room at various flow
Mach numbers M enables the derivation of scaling law after collapse of sound power
level curves as shown in Figure 45.
-50
Ls (dB)
-55
-60
-65
-70
-75
-1.2 -1 -0.8 -0.6 -0.4 -0.2 0
log(St)
Figure 45: Dimensionless spectrum LS for the straight pipe case, formulation used:
This section addresses the noise generated by flow noise for the rig (Figure 9) terminated
with elements such as oblique cuts, diffusers and bends. The methodology employed
consists of systematically comparing the element flow noise to the straight pipe flow
noise. Scaling laws are derived when additional flow noise of more than 3 dB is
observed between the element and the straight pipe.
82
Oblique cuts
The case corresponding to the most pronounced angle (60) is studied as it was
observed to be the one creating the highest sound power level. Measured flow noise
sound power levels are compared to the results for a straight pipe termination in Figure
46.
75
Straight pipe
70
65
M=0.19
Lw (dB)
60
Oblique cut 60
55
M=0.1
50
45
40
M=0.06
35
30
200
1000
2000
Frequency (Hz)
3000
4000
Figure 46: Flow induced sound power levels for oblique cuts.
83
Diffusers
Flow generated noise for diffusers are now presented and compared with the reference
case, i.e., the straight pipe termination.
80
Lw (dB)
70
M=0.05
Straight pipe
1
2.5
5
7.5
60
50
40
Straight pipe
30
80
Lw (dB)
70
M=0.09
Straight pipe
1
2.5
5
7.5
60
50
Straight pipe
40
30
80
Lw (dB)
70
60
Straight pipe
50
Straight pipe
1
2.5
5
7.5
40
30
84
The sound generation by flow for diffusers is of dipole type in free field. It means that
the sound power generated by the flow is proportional to its velocity U6 since the noise
generation occurs mostly at the opening. At low frequencies, a noticeable deviation
from the reference case can be observed for all investigated cases. The larger the
diffuser angle, the greater the sound power level generated by the flow. It is noted that
the diffuser lip circumferential length increases with the diffuser angle since the
diffusers have the same length. Consequently, flow separation and flow induced sound
power at the lips is most pronounced for diffusers with greater angle and same length
although the outlet airflow speed is reduced. The very small difference in the flow
induced sound power level LW between the diffuser with an angle of 1 and a straight
pipe shown in Figure 47 indicates that this diffuser may be regarded as a straight pipe in
terms of flow induced noise. Dimensionless spectra LS obtained for the diffuser cases
according to the formulation W = 0U 3 D 2 M CL2 F ( St ) are shown in Figure 48.
-25
=7.5
-30
Ls (dB)
-35
=5
-40
=2.5
-45
-50
-55
-60
-0.6
-0.4
-0.2
0
log(St)
0.2
0.4
Figure 48: Dimensionless spectra for the diffuser terminations, formulation used:
The collapse of data shown in Figure 48 for each angle was obtained for a velocity
dependence of U6 (outlet velocity, calculated from Pitot-tube measurement inside the rig
and outlet cross-section) corresponding to =3 in W = 0U 3 D 2 M CL2 F ( St ) ).
85
45 bends
Flow noise sound power level results for 45 bends are shown in Figure 49.
80
Lw (dB)
70
M=0.05
L=0
L=1D
L=2D
L=5D
60
50
40
Straight pipe
30
80
Lw (dB)
70
M=0.09
L=0
L=1D
L=2D
L=5D
60
50
Straight pipe
40
30
80
Lw (dB)
70
60
Straight pipe
50
L=0
L=1D
L=2D
L=5D
40
30
86
All bend elements behave as a straight pipe in the high frequency region. It is seen from
Figure 49 that the greater the flow Mach number, the smaller the deviation compared to
the results for a straight pipe. It is due to the flow noise at the opening dominating the
internal flow induced noise at the bend. Coupling occurs between the sound sources at
the bend and at the opening when the distance separating them is short (Nygrd (2000)).
The exception to this observation is for the bend with the longest upstream length
(L=5D). For a length of L=5D, the internal flow noise generated is greater than the
noise generation at the opening. The scaling laws for the 45 bends are shown in Figure
50. It shows more scattering for the shortest upstream lengths (L=1D and L=2D), due to
the mixed flow processes at the bend and at the opening. These two processes (internal
flow noise, flow noise at the opening) are described by two different velocity exponents.
The change in the velocity exponent is due to the presence of pipe walls (boundary
conditions) for the internal noise mechanism (Nelson & Morfey (1981)). No significant
difference between the shortest 45 bend (L=0) and the straight pipe in terms of flow
noise is observed, thus the results are not presented.
-10
-35
-20
L=5D
-30
-45
Ls (dB)
Ls (dB)
-40
-50
L=1D
-55
-40
-50
-60
-60
-70
L=2D
-80
-65
-0.5
-0.4
-0.3
-0.2 -0.1
log(St)
0.1
-0.3
-0.2
-0.1
0
0.1
Log(St)
0.2
0.3
0.4
The dimensionless spectra shown in Figure 50 have been obtained for a velocity
dependence of U4.
87
90 bends
Experimental data for flow generated sound for 90 smooth bends are given in Figure
51 together with the derived dimensionless spectra for the two 90 bends investigated
(Figure 52).
80
Lw (dB)
70
M=0.05
Straight pipe
R/D=1.6
R/D=2.5
60
50
40
Straight pipe
30
80
Lw (dB)
70
M=0.09
Straight pipe
R/D=1.6
R/D=2.5
60
50
Straight pipe
40
30
80
M=0.19
Lw (dB)
70
60
Straight pipe
50
40
30
Straight pipe
R/D=1.6
R/D=2.5
500 1000 1500 2000 2500 3000 3500 4000
Frequency
88
-30
-35
Ls (dB)
-40
R/D=1.6
-45
-50
R/D=2.5
-55
-60
-0.4
-0.3
-0.2
-0.1
0
0.1
Log(St)
0.2
0.3
0.4
Figure 52: Dimensionless spectra for the 90 bend cases, formulation used:
The bend with the largest curvature (R/D=1.6) generates the highest sound power level.
The dimensionless spectra (Figure 52) for the flow sound power generated in these bends
indicate that the sound generation is proportional to the airflow speed U4. The velocity
exponent =1 for the bends differs from that of the diffusers, i.e., =3. It is due to the
flow sound generation process, which for the diffusers mainly occurs at the opening
(free field) whereas it occurs inside the pipe (enclosed field) for the bends.
5.2.3 Scaling laws review
Figure 53 shows the scalings laws for diffusers and the straight pipe.
Nygrd (2000) has developed scaling laws for other type of bends. However, the
formulation based on Nelson & Morfey (1981) used in his derivation of dimensionless
spectra is slightly different from the procedure taken throughout this document and
needs to be converted. The conversion is given in (5-12). Complete sets of data from
Nygrd (2000) and our results are shown in Figure 53 together with the formulation
used to derive the dimensionless spectra LS.
89
-30
Diffuser 7.5
-35
-40
Ls (dB)
Diffuser 5
-45
Diffuser 2.5
-50
-55
-60
-65
Straight pipe
-0.3
-0.2
-0.1
0
0.1
log(St)
0.2
0.3
0.4
Figure 53: Complete sets of dimensionless spectra obtained from our investigation on diffusers.
-40
Mitre bend (Nygrd)
-45
Ls (dB)
-50
-55
90 bend R/D=1.6
-60
-65
90 bend R/D=2.5
-70
45 bend L=5D
-75
-0.4
-0.2
0.2
0.4
0.6
0.8
log(St)
Figure 54: Complete sets of dimensionless spectra obtained from our investigation on diffusers
and Nygrd (2000) after conversion to the following formulation excluding the pressure loss
coefficient CL: W = 0U 3 D 2 M F (St ) .
90
The semi-empirical laws together with coefficients (Table 3) obtained from Figure 53
and Figure 54 and a line fitting procedure are given in (5-13).
Ls = C1 [log (St )] + C 2 [log (St )] + C3 [log (St )] + C 4 [log (St )] + C5 [log (St )] + C6 . (5-13)
5
C1
C2
C3
C4
C5
C6
Straight pipe
-68
-44
28
12
10
-59
Diffuser 2.5
-82
25
33
-19
10
-48
Diffuser 5
34
11
-34
-28
15
42
Diffuser 7.5
19
19
-24
-33
16
-34
45 bend L=5D
100
73
-110
-49
-61
90 bend R/D=2.5
-420
-28
-63
-18
-61
90 bend R/D=1.6
310
-170
-120
-3
-2
-58
-150
-6
47
10
-18
-50
-27
10
-7.5
-54
91
5.3
Summary
An experimental investigation has been conducted on the noise created by airflow. The
experimental technique consists of a rotating microphone located in a reverberant room.
Thus, flow induced sound power levels have been measured from a flow rig terminating
with different geometries (oblique cuts, diffusers, bends) and at various flow Mach
number M <0.5. A preliminary investigation on a straight pipe termination has shown
that the dominant noise creation process occurs at the lips. Flow noise is also created
inside the pipe. However, this internal flow noise is of lesser magnitude compared to the
noise produced at the lips or externally for the Mach numbers studied here.
The flow noise sound power levels determined from the geometries have been
compared to the straight pipe case and derivations of scaling laws based on these results
have been obtained. The results demonstrate the different flow noise behaviours of the
elements investigated. It also indicates the relation between the airflow speed and the
noise produced due to the flow separation.
It has been found that the diffusers behave as the straight pipe in terms of the flow
induced noise process (same velocity exponent), i.e., the flow noise mechanism occur at
the outlet. Although, the flow speed is reduced due to the expansion of the diffuser, it
has been found that the flow noise at the diffuser outlet is increased compared to that of
a straight pipe. It is demonstrated that the diffuser elements behave as a straight pipe in
terms of flow induced noise. The sound generation mostly occurs at the opening/lips
(dipole sound mechanism). An increase of the angle of the diffuser (fixed length) results
in an augmented flow noise production, especially in the low frequency range (see
Figure 47). Moreover, for half angle greater than 4, flow separation also occurs
within the elements as the flow is detached from the wall. Due to the rather short length
of the diffuser, this noise creation inevitably interacts with the noise created at the lips,
though it is believed that the latter is of greater magnitude. On the other hand, the
oblique cut elements do not generate additional noise and can be treated as a straight
pipe. The bends require a special treatment as two flow noise processes can be
accounted for. Flow noise is created internally at the bend and externally at the lips. The
internal process is of main interest and a correction technique has been applied to
retrieve the flow noise at the bend from external measurement. Semi-empirical laws to
calculate flow noise sound power levels have been derived for all elements tested.
92
93
internal combustion engine. Reflection coefficient models for the various tailpipes
studied combining Munts theory and transfer matrix calculations have been tested
against measurements performed induct with flush mounted microphones. For the
diffusers an interesting deviation from the straight pipe behaviour was found, since the
magnitude of the reflection coefficient is larger than 1 in the low frequency limit. This
effect gives a significant reduction up to 10 dB of tailpipe resonances. It has been found
that the phase difference at various points along the cut needs to be modelled for the
oblique cut elements. Bends behave as a straight pipe regarding the internal sound
propagation.
In terms of flow noise, scaling laws formulation based on Nelson & Morfey (1981) for
calculation of flow induced sound power as function of frequency and characteristic
dimension, were obtained in a similar fashion as for the straight pipe. A sixth power
velocity dependence was found for the diffuser cases. Bends scale with U4. No
additional noise is created by the oblique cuts.
Regarding future research, one effect that could be of importance in IC-engine intake
and exhaust systems is non-linearities since the pressure amplitude of the fundamental
engine harmonics can reach 160-170 dB. Both the diffusers and bends studied are
elements for which non-linearities can play a role (Dequand and al. (2002)).
Temperature gradients are known to be of little effect.
A more refined analysis on the sound propagation in oblique cuts is needed to model the
role of the sharp cut on the acoustics of such elements.
A complete form of the theory of Munt (1977, 1990) gives also the far-field directivity
pattern. However, the coded model that was used in this document concerned only the
reflection coefficient at the opening from which impedance and end correction were
calculated. A refined coding giving the far-field behaviour, compared with the sound
transmission measurement obtained here would also give the sound absorption
magnitude in the near field. In other words, prediction of acoustical energy losses that
occur in the flow could be also performed using Munts model.
94
Oblique cuts
Deviations from the straight pipe at 2.5, 5, 7.5: Flow noise sound power
Bends
level at opening U 6
45 bend(L=0): Equivalent to a
straight pipe
45 bend(L=5D), 90bends (R/D=1.6,
2.5): Flow noise sound power level at
the bend U 4
95
96
7. References
1. bom M., 1989, Studies of sound propagation in ducts and mufflers, Doctoral Thesis,
Department of Aeronautical and Vehicle Engineering, Royal Institute of Technology,
Stockholm, Sweden.
2. bom M., Bodn H., 1988, Error analysis of two-microphone measurements in ducts with
flow, Journal of the Acoustical Society of America, 83, 2429-2438.
3. Allam S., 2004, Acoustic modelling and testing of advanced exhaust system components for
automotive engines, Doctoral Thesis, Department of Aeronautical and Vehicle Engineering,
Royal Institute of Technology, Stockholm, Sweden.
4. Baxa D. E., 1989, Noise control in combustion engines, Krieger Pub Co.
5. Bechert D. W., 1980, Sound absorption caused by vorticity shedding, demonstrated with a jet
flow, Journal of Sound and Vibration, 70, 389-405.
6. Bierkens J., 2002, Calculation of the reflection coefficient of a jet pipe with subsonic jet flow,
Master Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of
Technology, Stockholm, Sweden.
7. Blevins R. D., 1984, Applied fluid dynamics handbook, New York : Van Nostrand Reinhold.
8. Bodn H., 1989, Characterization of fluid machines as sources of fluid borne noise, Doctoral
Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of Technology,
Stockholm, Sweden.
9. Bodn H., Carlsson U., Glav R., Wallin H. P., bom M., 2001, Ljud och vibration, Norstedts
Tryckeri AB, Stockholm.
10. Boij S., 2003, Acoustic scattering in ducts and influence of flow coupling, Doctoral Thesis,
Department of Aeronautical and Vehicle Engineering, Royal Institute of Technology,
Stockholm, Sweden.
97
11. Callow G.D., Peat K.S., 1988, Insertion loss of engine intake and exhaust silencers, IMechE
C19/88, 39-46.
12. Cargill A. M., 1982, Low frequency acoustic radiation from a jet pipe- a second order
theory, Journal of Sound and Vibration, 83, 339-354.
13. Davies P.O.A.L., 1988, Practical flow duct acoustics, Journal of Sound and Vibration,
124(1), 91-115.
14. Davies P.O.A.L., Coelho J. L. B., Bhattacharya M., 1980, Reflection coefficients for an
unflanged pipe with flow, Journal of Sound and Vibration, 72, 543-546.
15. Dequand S., van Lier L., Hirschberg A., Huijnen J., 2002, Aeroacoustics response of
diffusers and bends: comparison of experiments with quasi-steady incompressible flow models,
Journal of Fluids and Structures, 16, 957-969.
16. Dokumaci E., !998, On transmission of sound in circular and rectangular narrow pipes with
superimposed mean flow, Journal of Sound and Vibration, 210, 375-389.
17. Ducret F., bom M., 2004, Aeroacoustics behaviour of different exhaust pipe geometries,
Eleventh International Congress on Sound and Vibration, St. Petersburg, Russia.
18. Ducret F., 2005, Aeroacoustics behaviour of a straight pipe, Twelfth International Congress
on Sound and Vibration, Lisbon, Portugal.
19. Gijrath H., Nygrd S., bom M., 2001, Modelling of flow generated sound in ducts, Eight
International Congress on Sound and Vibration, Hong-Kong, China.
20. Glav R., 1994, On acoustic modelling of silencers, Doctoral Thesis, Department of
Aeronautical and Vehicle Engineering, Royal Institute of Technology, Stockholm, Sweden.
21. Goldstein M., 1976, Aeroacoustics, McGraw-Hill, New York.
22. Hirschberg A., Bruggeman J. C., Wijnands A. P. J., Morgenstern M., 1988, The whistler
nozzle and horn as aero-acoustic sound sources in pipe systems, Proc. Inst. Acoust., 10, 701708.
98
23. Howe M. S., 1980, The dissipation of sound at an edge, Journal of Sound and Vibration, 70,
407-411.
24. Howe M. S., 1995, The damping of sound by turbulent shear layers, Journal of the
Acoustical Society of America, 98, 1723-1730.
25. Huijnen J.H., 1998, Aero-acoustics of a bend: quasi-stationary models, Technische
universiteit eindhoven, Report R1476A.
26. Ingard U., Singhal V. K., 1974, Sound attenuation in turbulent pipe flow, Journal of the
Acoustical Society of America, 55, 535-538.
27. int panhuis P., 2003, Calculations of the sound pressure reflection coefficient and the
acoustic end correction of a semi-infinite pipe issuing a subsonic cold or hot jet with co-flow,
Master Thesis, Department of Aeronautical and Vehicle Engineering, Royal Institute of
Technology, Stockholm, Sweden.
28. ISO 3741, 1994, Acoustics: Determination of sound power levels of noise sources using
sound pressure, precision methods for reverberation rooms.
29. ISO 3747, 1998, Acoustics: Determining the sound power level of noise sources from sound
pressure measurement.
30. Kirchoff G., 1868, Uber den einfluss der wrmteleitung in einem gase auf die
schallbewegung, Pogg. Ann., 134(6), 177-193.
31. Kuhn G.F., Morfey C.L., 1976, Noise due to fully developed turbulent flow exhausting from
straight and bend pipes, Journal of Sound and Vibration, 44, 27-35.
32. Levine, H., Schwinger H., 1948, On the radiation of sound from an unflanged circular pipe,
Physical Review, 73, 383-406.
33. Morfey C.L., 1971, Sound transmission and generation in ducts with flow, Journal of Sound
and Vibration, 14(1), 37-55.
99
34. Munt R.M., 1977, The interaction of sound with a subsonic jet issuing from a semi-infinite
cylindrical pipe, Journal of Fluid Mechanics, 83, 609-61040.
35. Munt R. M., 1990, Acoustic transmission properties of a jet pipe with subsonic jet flow: I.
the cold jet reflection coefficient, Journal of Sound and Vibration, 142, 413-436.
36. Nelson P.A., Morfey C.L., 1981, Aerodynamic sound production in low speed flow ducts,
Journal of Sound and Vibration, 79, 263-289.
37. Nygrd S, 2000, Modelling of low frequency sound in duct networks, Licenciate Thesis, ,
Department of Aeronautical and Vehicle engineering, Royal Institute of Technology,
Stockholm, Sweden, TRITA-FKT 2000:57.
38. Oldham D. J., Ukpoho A. U., 1990, A pressure-based technique for predicting regenerated
noise levels in ventilation systems, Journal of Sound and Vibration, 140, 259-272.
39. Peters M.C.A.M., Hirschberg A., Reijnen A.J., Wijnands A.P.J, 1993, Damping and
reflection coefficient measurements for an open pipe at low Mach and Low Helmholtz numbers,
Journal of Fluid Mechanics 256, 499-534.
40. Pierce A.D., 1989, Acoustics: an introduction to its physical principles and applications,
Acoustical Society of America, Mc Graw Hill Inc., New York.
41. Rienstra S.W., Hirschberg A., 2004, An introduction to acoustics, Report IWDE 01-03 ,
Eindhoven University of Technology, can be downloaded at
http://www.win.tue.nl/%7Esjoerdr/papers/boek.pdf
42.
Ronneberger
D.,
1975,
Genaue
messung
der
schalldmpfung
und
der
100
44. Schlinker R.H., 1979, The transmission of acoustic plane waves as a jet exhaust, Journal of
Aircraft, 16(3).
45. Tijdeman H., 1975, On the propagation of sound in cylindrical tubes, Journal of Sound and
Vibration, 39, 1-33.
46. van Lier L.J., 1999, An experimental study of the dynamic response of diffusers with
application to gas transport system, Technische universiteit Eindhoven, Report R-1493-A.
47. van Lier L., Dequand S., Hirschberg A., Gorter J., 2001, Aeroacoustics of diffusers: An
experimental study of typical industrial diffusers at Reynolds numbers of O(105), Journal of the
Acoustical Society of America, 109, 108-115.
48.
VDI3733,1996,
Noise
at
pipes,
VDI
manual
noise
reduction
guidelines,
101
Paper III. Linear and non linear acoustic regimes of perforated plates
Abstract
The paper treats the sound transmission through perforated sample. Micro-perforated
elements are being tested as a substitute for hazardous porous material used for
automotive application. An experimental investigation is performed on the non linearity
of samples perforated with different types of apertures (hole, square, rectangular). The
parameter of interest is the specific acoustic impedance defined as the ratio of the
pressure drop across the thickness of the sample to the particle velocity. Non linear
resistive results are used to obtain the perforation area ratio (porosity) of the samples
tested. A linear analytical model for the flow impedance is formulated. It is composed
of a series of uncoupled internal and external resistance and reactance terms. The model
is confronted with the results at low particle velocity (linear regime).
Table of Contents
1. Introduction .................................................................................................................. 1
2. Modelling the acoustics linear regime of perforated plates.......................................... 5
2.1.1
2.1.2
2.1.3
4.1.1
4.1.2
Samples................................................................................................... 15
4.1.3
5.2
General discussion.......................................................................................... 32
5.3
5.4
6. Conclusion .................................................................................................................. 41
7. References .................................................................................................................. 43
1. Introduction
Perforated panels have been used in the automotive industry for decades not primarily
as a mean to attenuate sound since the perforations are too large (>1 mm) to provide
substantial viscous-thermal dissipation but to guide the airflow to ensure low back
pressure, a parameter of paramount importance for the design of silencing systems. A
reactive silencer is based on the impedance mismatch due to the expansion chamber.
This results in the reflection of incoming waves. An additional acoustic treatment is to
line the inside wall of this chamber with porous materials to create a resistive
(dissipative) type of silencer element. The separation between the silencer core and the
damping cavity is normally provided by a perforated tube. It has been demonstrated that
for a perforation area ratio of more than 30%, the panel is acoustically transparent.
The use of micro-perforated (apertures of submillimetre size) material ensures enough
losses by itself so that no bulky absorbent material is required. The other advantage over
traditional porous/fibre type of materials is the higher attenuation efficiency at low
frequency and the reduction of the hazardous risks associated with the handling of
brittle material.
Two mechanisms are responsible for the dissipation of the acoustic energy by microperforated panels. The visco-thermal boundary layers formed along the inner walls of
the aperture and caused by the fluid interacts with the acoustic waves. This interaction
cannot be neglected for apertures of small transverse dimension and causes the
acoustical energy to be converted into heat. Another mechanism is the acoustic particle
jet formation (vorticity) at the opening. The acoustic particle velocity is increased
sharply as the wave is squeezed into the minute perforations. A jet is formed as the
wave exits the other side of the perforation. For high sound amplitude, this results in the
formation of vorticity at the opening. This non-linear phenomenon enhances the
dissipation of acoustical energy.
It is known (Ingard (1953), Morfey (1969)) that the acoustics of apertures depends on
their geometries. For thin panel and at low frequency, the sound transmission through
orifices depends mostly on the perimeter of the opening (Morfey (1969)).
Analytical and experimental investigations have been performed by various authors on
the acoustics of panels composed of a single circular hole (Ingard (1953), Ingard &
1
Ising (1967)) and a single rectangular aperture (Sivian (1935)). Maa (1998) details a
unified formulation for the acoustic impedance of panels with multiple micro-circular
holes. Maas analytical model treats the internal and external resistances and reactances.
No similar model exists for panels composed of multiple rectangular apertures.
Attempts (Gomperts (1964, 1965), Gomperts & Kihlman (1967), Ingerslev & Nielsen
(1944)) by building acousticians have been performed on modelling the scattering of
sound on cracks in walls or opening around door frames. These models regard the slits
as infinitely long and do not treat the viscous dissipation inside and on the edges of the
aperture. The cases presented in this paper are on the other hand considered of finite
dimensions and the internal and external acoustics behaviours including viscous losses
are detailed. Therefore, the building acoustics models are inadequate for the present
configurations.
A more closer approach on modelling the acoustics of perforated slitted panels has been
carried out by Randeberg (2000, 2002). He describes an analytical model coupled with a
finite element model for calculation of the absorption coefficient of double slitted
perforated panels.
Based on Maas method (Maa (1998)) for micro-circular holes and Randebergs work
(Randeberg (2000, 20002)) on slitted plates, an analytical model for the sound
transmission through a single panel composed of rectangular apertures is derived in this
paper and each of the terms are discussed. The aim of this derivation is to allow easy
and flexible calculations of perforated slitted panel properties to attenuate sound.
One of the main assumptions made when modelling the perforated elements is that the
apertures do not interact. Therefore, it is convenient to derive the theory for the acoustic
impedance of one single aperture and use the perforation area ratio (porosity) to model
the entire panel. The normal incident impedance Z = p u over the panel is defined as
Z = [(1 + R ) (1 R )] 0 c0 = 0 c0
coefficient R = p p + .
The
where
dimensionless
is
normal
the
pressure
impedance
is
reflection
defined
by
Acustimet panel is composed of small slits obtained after punching aluminium or steel
plates. This manufacture process is relatively fast and inexpensive to perform. The
micro-perforated Acustimet product has been already implemented with success in
cavity to dampen the acoustics modes associated with standing waves. A new area of
interest is the use of this material for exhaust systems. The potential for silencers made
of this material is still not well known and the effects of high amplitudes and flow on
the behaviour should be investigated.
A closer look at the Acustimec product reveals the presence of minute apertures of
geometry closer to rectangular than circular orifices. An analytical model based on
panels composed of rectangular apertures is therefore required to predict the acoustics
of such panels.
Previous works (Ingard (1953), Maa (1998)) have relied mostly upon the measurement
on only one side of the sample. The experimental apparatus consists usually of one
measuring section (impedance tube) terminated by the sample and a rigid cavity
(resonator). Microphones mounted on one section allow determination of the reflection
and absorption coefficients. The techniques presented in this paper are related to a rig
comprised of two measuring sections. The sample is mounted between these sections
with anechoic terminations and measurements are performed on both sides of the
sample. Thus, the sample can be characterised more exhaustively.
Non-linear regime implies that the acoustic properties, for example, the flow
impedance, are dependent of the acoustic particle velocity (amplitude of the acoustic
source). Thus, non-linear parameters are required for prediction of sample properties at
high sound intensity. It has been demonstrated (Maa (1998)) that there is not a sharp
distinction between the linear and non-linear regimes.
(2-1)
where p is the acoustic pressure drop through the thickness t of the orifice and v is the
interior acoustic particle velocity. This is related to the external velocity u A given in
equation (4-11). The impedance is composed of a resistance term R and reactance term
X. The plate is assumed to be thin so the particle velocity v is assumed to be the same
on both sides.
The following section sets out the analytical expressions for the various analytical terms
for the acoustic impedance of circular and rectangular orifices with geometries depicted
in Figure 1.
Internal terms
Internal viscous losses
Internal propagation of acoustic waves through narrow apertures causes the acoustical
energy to be converted into other forms of energy. The main sound dissipation
mechanism encountered in sound propagation through small apertures is of viscous
nature. Thermal effect corresponding to the heat conduction between the fluid and the
wall is usually neglected.
Allard (1993) derives the complete internal acoustic impedance Zi,ra of a single
rectangular aperture. The internal resistive term Ri,ra for rectangular geometry is taken
as the real part of the expression for Zi,ra.
Ri ,ra
tanh w j 0
= Re j 0t 1
w j 0
(2-2)
Internal reactance
The internal reactance Xi,ra for single rectangular aperture can be extracted by taking the
imaginary part of the complete internal impedance formulation Zi,ra (see equation (2-5))
derived by Allard (1993).
X i ,ra
tanh w j 0
= Im j 0t 1
w j 0
(2-3)
X i ,ra = j 0 t
(2-4)
Internal impedance
The complete internal impedance Zi,ra for a slit orifice with no thermal effect has been
calculated by Allard (1993) and Cremer & Mller (1982). It is calculated from the
effective density eff of the air inside the slit following Kirchoffs method. It is assumed
that the width w is greater than height h but less than the acoustic wavelength. The
thermal losses are neglected. It is expressed as
Z i ,ra
tanh w j 0
2
= j eff t = j 0t 1
w j 0
(2-5)
External terms
External friction resistance and radiation resistance
Under the condition of the non presence of porous materials or screens, the acoustic
energy is mostly dissipated through viscous effects and heat conduction on the surfaces
of the aperture.
The external resistance Re corresponds to these losses caused at the opening due to the
friction against the opening surfaces. Ingard (1953) obtains
Re ,ca = RS = (1 2 ) 2 0 .
(2-6)
This formulation is valid for the case of an oscillating flow over a surface. This formula
must be modified for the effects of the edges of the opening. Ingard (1953) suggests
based on experiments a value of 4Rs. Maa (1998) argues that this value is too high and
sets it to Rs in his model for micro-circular holes. This is probably due to the fact that he
uses Crandalls internal resistance term (Crandall (1927)) whereas Ingard (1953) treats
the internal losses with an approximate form. Randeberg (2000, 2002) uses a factor of 2
together with the internal resistance predicted by Allard (1993). This model will also be
used here, i.e.,
Re ,ra = 2 0 = 2 RS .
(2-7)
Sivian (1935) calculates the air mass reactance on a rectangular aperture based on the
following assumptions:
-The particle velocity v normal to the plane of the aperture is constant across the
orifice.
-The aperture transverse dimensions (width w, height h) are small compared to the
wavelength (kw<<1, kh<<1).
-The aperture is located in an infinite baffle.
These assumptions allow him to write a simple and easy to calculate velocity potential
based on v . The end correction for one side of a rectangular aperture is then given by
(w 2 )3 + (h 2 )3 (w 2 )2 + (h 2 )2 3 / 2 +
(w 2 ) + (w 2 )2 + (h 2 )2 1 2
(h 2 ) ln
.
h2
ra =
3 (w 2 )(h 2 ) 3(w 2 )(h 2 )
(h 2 ) + (w 2 )2 + (h 2 )2 2
+ (w 2 ) ln
2
w
(2-8)
Our analytical model for rectangular apertures includes the end correction on both sides.
Mass attached reactance
The external mass attached reactance Xe for one side of the aperture derived from the
end correction term is of the form
X e = 0 .
(2-9)
It is now time to assemble all the previous terms into a single expression for the acoustic
impedance Zra of an orifice of rectangular geometry at low sound intensity.
Z ra = Z i ,ra + Z e ,ra = 2 Re ,ra + Z i ,ra + 2 jX e ,ra
1
w j 0
tanh
2
t 1
+
w j 0
3/ 2
3
3
2
2
(w 2 ) + (h 2 ) (w 2 ) + (h 2 )
.
+
= 2 0 + j 0
(w 2 ) + (w 2 )2 + (h 2 )2 1 2
(h 2 ) ln
h
2
3 (w 2 )(h 2 ) 3(w 2 )(h 2 )
1
(h 2 ) + (w 2 )2 + (h 2 )2 2
+ (w 2 ) ln
w2
(2-10)
Having reviewed the expression (2-10) for the acoustic impedance Zra for a single
rectangular orifice, the next step is to compute the overall acoustic impedance for the
panel depicted in Figure 3.
(b)
Figure 3: Schematic of perforated samples with n apertures and cross section S=L1L2.
The total panel acoustic impedance calculated for panels as depicted in Figure 3 and
normalised with respect to 0c0 is given by
(2-11)
= nhw S ,
(2-12)
10
11
R = 0 v (1 2 ) (1 q1 ) ,
(3-1)
where is the perforation area ratio (porosity) and q1 is the fraction of the excitation
pressure corresponding to the acoustic pressure inside the hole.
Maa (1994) suggests a modified version of this for the case of micro-perforated plates
with small porosity and expresses the normalised non linear resistance term for a
complete perforated panel as
r = v (c0 ) = u A ( 2c0 ) ,
where u A is the acoustic particle velocity on the inlet sample surface.
12
(3-2)
4. Experimental apparatus
4.1
In this section, we describe the experimental apparatus and techniques used to measure
the sample normalised acoustic impedance S and the inlet sample particle velocity
u A and the normalised acoustic impedance S based on the interior particle velocity v .
The rig shown in Figure 6 and Figure 5 consists of two identical sections. Positions 1 and
2 are part of the upstream1 section whereas positions 3 and 4 are referred to the
downstream1 section.
The two sections are clamped and tightened together through flanges. To prevent air
leakages, rubber joints are installed on each section at the location of the sample.
The walls of the duct are perfectly rigid so that no vibration occurs. The duct is
terminated on both sides with porous material to reduce standing waves inside the duct.
Mounts for microphones have been manufactured to facilitate the change of position.
The cut-off frequency for this square duct is given by f c = c0 2 L = 2450 Hz where c0 is
the speed of sound (343 m s 1 ) and L is the duct transverse dimension (70 mm). Plane
wave propagation occurs below the cut-off frequency. No higher order modes are
investigated.
The upstream and downstream sections refer to the rectangular tubes located on both sides of the
13
14
Figure 6: Layout of the experimental rig for sample testing.
4.1.2 Samples
Table 1 shows the samples investigated together with their dimensions in mm. The total
Acustimet
15
The main methodology consists of determining the sound fields on either side of the
sample. The difference between them is attributed to the properties of the material. No
direct measurement is made on the sample. The requirement is to map the propagation
of acoustic waves in both sections of the duct (Figure 7).
In order to achieve this, two microphones are used simultaneously. Frequency response
(transfer functions) measurements are carried out between the two microphones. The
data acquisition system and signal to the loudspeaker was provided by a Brel & Kjr
PULSE system. The PULSE software programming language necessitates to derive the
real and imaginary parts of the parameters investigated. The drawback of using of two
microphones simultaneously is that a phase and magnitude calibration between the
microphones is needed.
The experimental technique requires a separate phase and magnitude calibration for the
two microphones. This calibration procedure in this case consists of placing the two
microphones through a coupler mounted on a calibrator producing random noise. By
measuring the sound intensity, one can adjust the phase difference within the PULSE
software to compensate for any differences. The calibration transfer function H21
16
cal
between the two microphones is saved. The index 1 and 2 in the calibration procedure
refers to the microphone number (1 and 2) whereas it indicates the duct position number
(1 through 4) when they are mounted on the duct. The calibration saved transfer
function H21 cal is combined with the measured frequency response H21 meas to calculate
the true frequency response H21 true as follows
) [ (
) (
) (
)] [Re(H ) + Im(H )] ,
) [ (
) (
) (
2
21cal
2
21cal
2
21cal
(4-1)
2
21cal
From now on, the true frequency response for the automation technique is simply
referred to H21.
Overview of the technique
location 4) is calculated from equation (4-1) using the calibration file. The calculation of
the pressure reflection coefficient R4 at position 4 for the downstream condition is
further performed. The real and imaginary parts for R4 are given as
H 43 + e jks34
R4 =
,
H 43 e jks34
Im(R4 ) =
(4-2)
RB = R4e 2 j (k ( s34 + l3 s )) ,
Re(RB ) = Re(R4 ) cos (2k (s34 + l3s )) + Im(R4 ) sin(2k (s34 + l3s )),
(4-3)
Im(RB ) = Im(R4 ) cos (2k (s34 + l3s )) Re(R4 ) sin(2k (s34 + l3s )).
The normalised normal impedance on sample surface B is thus
B =
1 + RB
,
1 RB
1 (Re(RB )) (Im(RB ))
,
Re( B ) =
(1 Re(RB ))2 + (Im(RB ))2
2
Im( B ) =
2 Im(RB )
.
(1 Re(RB ))2 + (Im(RB ))2
18
(4-4)
Upstream properties
H 21 + e jks12
,
H 21 e jks12
Im(R2 ) =
(4-5)
(4-6)
A =
1 + RA
,
1 RA
19
(4-7)
Re( A ) =
1 (Re(RA )) (Im(RA ))
,
(1 Re(RA ))2 + (Im(RA ))2
Im( A ) =
2 Im(RA )
.
(1 Re(RA ))2 + (Im(RA ))2
The normal impedance A and saved normal impedance B are further combined to
derive the sample normalised acoustic impedance S given as
S =
1 + R A 1 + RB
= rs + jx s .
= ( A B ) =
u A 0 c0
1 R A 1 RB
p
(4-8)
An extra measurement is required on the upstream section. One of the aims of this
investigation is to estimate the dependence of the level of excitation on the acoustic
impedance of the sample mounted between the two sections. Therefore, one also needs
to measure the particle velocity u A on the surface A (see Figure 7) of the sample using a
wave decomposition technique.
The expressions used are
u A =
Re(u A ) =
Im (u A ) =
[Re(H ) Re(H )]
2 A+
2 A
0c0
[Im(H ) Im(H )]
2 A+
2 A
0c0
S22
,
K2
S22
,
K2
Re(H 2 A + ) =
Im (H 2 A + ) =
20
(4-9)
Re(H 2 A ) =
Im (H 2 A ) =
Consequently, it can be seen from equation (4-9) that the experimental derivation of the
particle velocity u A at the surface A requires the measurement of the transfer function
H21, the autospectrum S22 at position 2 (microphone 2) together with the magnitude
calibration K2 [V/Pa] of the microphone 2 located at position 2. For a thin sample, the
particle velocity is the same on either side, i.e., u A = u B .
Acoustic impedance based on the interior particle velocity v
The normalised acoustic impedance based on the interior particle velocity v is defined
as
S =
p
p
= S = rS + j x S
=
v 0 c0 (u A ) 0 c0
(4-10)
(4-11)
where u A corresponds to the averaged particle velocity over the inlet sample cross
sectional area given in equation (4-9) and is the porosity.
21
22
2.5
Normalised resistance
100 Hz
2
1.5
1
340 Hz
0.5
200 Hz
0
0
0.02
0.04
0.06
0.08
0.1
Peak particle velocity on the sample inlet surface (m/s)
0.12
2.2
2
Normalised reactance
1.8
340 Hz
1.6
1.4
1.2
200 Hz
1
0.8
0.6
100 Hz
0.4
0.2
0
0.02
0.04
0.06
0.08
0.1
Peak particle velocity on the sample inlet surface (m/s)
0.12
Figure 8: Normalised impedance S for the rectangular aperture (w=50 mm, h=1 mm,
=0.0102).
23
Normalised resistance
2.5
2
100 Hz
1.5
0.5
0
800 Hz
0.01
0.02
0.03
0.04
0.05
0.06
Peak particle velocity on the sample inlet surface (m/s)
0.07
2.4
2.2
Normalised reactance
800 Hz
1.8
1.6
1.4
400 Hz
1.2
1
0.8
0.6
0.4
0
100 Hz
0.01
0.02
0.03
0.04
0.05
Peak particle velocity on the sample inlet surface
0.06
0.07
Figure 9: Normalised impedance S for a plate with multiple circular holes (d=0.4 mm, t=0.6
mm =0.0056).
24
0.8
Normalised resistance
0.7
0.6
100 Hz
0.5
300 Hz
0.4
0.3
0.2 700 Hz
0.1
0.05
0.1
0.15
0.2
0.25
0.3
Peak particle velocity on the sample inlet surface (m/s)
0.35
0.4
0.35
Normalised reactance
0.3
700 Hz
0.25
0.2
0.15
300 Hz
100 Hz
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
Peak particle velocity on the sample inlet surface (m/s)
0.35
Figure 10: Normalised impedance S for a plate with multiple circular holes (d=0.5 mm, t=0.5
mm =0.041).
25
2.6
2.4
Normalised resistance
2.2
100 Hz
2
1.8
300 Hz
1.6
1.4
1.2
1
0.8
0
800 Hz
0.05
0.1
0.15
0.2
Peak particle velocity on the sample inlet surface (m/s)
0.25
0.7
0.6
800 Hz
Normalised reactance
0.5
0.4
0.3
0.2
300 Hz
0.1
100 Hz
0
0.05
0.1
0.15
0.2
Peak particle on the sample inlet surface (m/s)
0.25
26
3.5
100 Hz
Normalised resistance
2.5
300 Hz
1.5
0.5
0.005
0.01
0.015
0.02
Peak particle velocity on the sample inlet surface (m/s)
0.025
3.5
Normalised reactance
3
300 Hz
2.5
1.5
0.5
100 Hz
0.005
0.01
0.015
0.02
Peak particle velocity on the sample inlet surface (m/s)
0.025
Figure 12:Normalised impedance S for a plate with four rectangular apertures (w=5 mm, h=1
mm,=0.004).
27
5.1
Normalised resistance
based on the particle velocity inside the aperture
0.045
0.04
0.035
0.03
0.025
0.02
100 Hz
0.015
0.01
0.005
0
0
4
6
8
10
Peak particle velocity inside the aperture (m/s)
Normalised reactance
based on the particle velocity inside the aperture
0.018
12
0.016
0.014
0.012
100 Hz
0.01
0.008
0.006
0.004
0.002
0
0
4
6
8
Peak particle velocity inside the aperture (m/s)
10
12
Figure 13: Normalised impedance S based on the interior particle velocity v for plates with
single square and rectangular apertures.
28
Figure 14 shows the acoustic impedance based on the internal particle velocity v for
different excitation levels for plates with the same porosity and different single aperture
geometries.
It is immediately apparent from Figure 14 that the overall acoustic behaviours for both
regimes are not related to the type of geometry. The square, rectangular, hole apertures
all show a linear increase of the resistive rS and a drop of the reactance xS above a
certain transitional acoustic particle velocity vt .
It can be seen from Figure 14 that the linear and non linear normalised resistances rS
appear to be independent of the frequency. The normalised resistance is related mostly
to the porosity of the sample.
The normalised reactance xS is the same for the square and circular orifices with same
porosity. An elongation of the aperture results in a reduction of the end correction. The
transitional particle velocity vt increases with the frequency. Figure 13 indicates that an
elongated aperture causes the particle velocity vt to be reduced. This can be explained
on the basis that vorticity is more likely to be formed as the aperture narrows.
Figure 13 emphasises the point previously discussed concerning the transitional particle
velocity vt . The narrower the dimensions, the smaller the particle velocity vt . The
square and rectangular apertures (w=52 mm, h=2.5 mm) have the same porosity
implying the same averaged particle velocity u A over the entire sample cross section.
Comparison between these two geometries shows that an elongated aperture results in a
lower reactance. This is due to the end correction term, the external reactance being
nearly identical due to the same thickness t=1mm.
29
0.045
Normalised resistance
based on the particle velocity inside the aperture
0.04
w=11.4 mm, h=11.4 mm
0.035
d=12.9 mm
0.03
100 Hz
0.025
0.02
0.015
100 Hz
300 Hz
0.01
300 Hz
0.005
300 Hz
0
0
4
6
8
10
Peak particle velocity inside the aperture (m/s)
Normalised reactance
based on the particle velocity inside the aperture
0.05
14
300 Hz
0.045
12
0.04
d=12.9
mm
0.035
300 Hz
0.03
0.025
0.02
0.015
100 Hz
100 Hz
0.01
100 Hz
0.005
0
4
6
8
Peak particle velocity inside the aperture (m/s)
10
12
Figure 14: Normalised impedance S based on the interior particle velocity v for plates with
single square, rectangular and hole orifices ( =0.0265).
30
Multiple apertures
Figure 15 shows the impedance for a plate composed of the same slit dimension (w=5
Normalised resistance
based on the particle velocity inside the aperture
0.02
0.018
0.016
100 Hz
0.014
0.012
0.01
2 slits
4 slits
16 slits
0.008
0.006
0.004
0.002
0
0
3
4
5
6
7
8
Peak particle velocity inside the aperture (m/s)
10
-3
Normalised reactance
based on the particle velocity inside the aperture
5.5
x 10
2 slits
4 slits
16 slits
100 Hz
4.5
3.5
2.5
2
0
2
3
4
Peak particle velocity inside the aperture (m/s)
Figure 15: Single aperture normalised impedance S for plates with multiple rectangular
aperture (w=5 mm, h=1 mm).
31
Figure 15 demonstrates the effect of the coupling between adjacent rectangular apertures
as Ingard (1953) and Melling (1973) have discussed for circular orifices. An increase of
porosity results in a reduction of both the resistance and reactance linear results. It also
appears than the limit of 4% for the porosity previously assumed has been
overestimated. Coupling effects take place for lower porosity values in the case of
aperture of rectangular geometry.
5.2
General discussion
In view of the experimental results obtained, some empirical observations can be made.
The linear limit seems more or less independent of the frequency for the resistive term.
For the reactance the linear limit on the other hand is frequency dependent and increases
with the frequency. Commenting the non-linear regime now, the resistive term is
linearly proportional to the peak particle velocity increases and independent of the
frequency. The non-linear reactance is reduced as the particle velocity u A rises. From
Figure 14 it is clear that the transition between the linear and non-linear regime is
dependent of the geometry of the orifice. Non-linearity effects are triggered for lower
interior particle velocity vt for the rectangular aperture than for the square aperture.
This is due to the narrower dimension causing the acoustic flow to break down into
vorticity. This observation emphasises the argument of Maa (1994) that non-linearity is
an external phenomenon. Maa (1994) states that the internal impedance is independent
of the sound intensity.
5.3
Following Ingard (1953) and Maa (1998) treatments of the non linear resistive
impedance, a procedure is now discussed to measure the perforation ratio area of the
samples. At high (non-linear regime) sample inlet particle velocity u A , the normalised
resistance of the sample is observed (Figure 8-Figure 12) to be linearly related to the
particle velocity u A . Using a line fitting technique, the slope of the normalised
resistance as function of the particle velocity u A is obtained for the non-linear regime.
According to Maa (1998), the slope is equal to 1 ( 2 c0 ). Therefore, the perforation ratio
area is obtained from the measured non-linear normalised flow resistive term rS.
32
Results for the samples investigated are given in Table 2 and show good agreement with
the true porosity calculated from the dimensions of the orifice and the entire plate.
Samples
Measured porosity
Calculated porosity
0.001
0.002
0.0102
0.004
0.016
0.0265
0.0265
0.0267
(d=0.4 mm)
Multiholed plate 2
(d=0.5 mm)
Sontech Acustimet microperforated sample
0.0056
0.041
5.4
Plate modelling
The plain straight line corresponds to the calculated value for the linear regime based on
equation (2-10).
33
single aperture
1.8
1.6
100 Hz
Normalised resistance
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
0.05
0.1
0.15
0.2
0.25
Peak particle velocity on the sample inlet surface (m/s)
1
0.9
100 Hz
Normalised reactance
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.05
0.1
0.15
0.2
0.25
Peak particle velocity on the sample inlet surface (m/s)
0.3
Figure 16: Normalised impedance S for a plate with one rectangular aperture (w=52 mm,
h=2.5 mm, =0.0265).
34
Normalised resistance
2.5
100 Hz
1.5
0.5
0
0
0.05
Peak particle velocity on the sample inlet surface (m/s)
0.1
1.6
1.4
Normalised reactance
1.2
100 Hz
1
0.8
0.6
0.4
0.2
0.02
0.04
0.06
0.08
0.1
Peak particle velocity on the sample inlet surface (m/s)
0.12
Figure 17: Normalised impedance S for a plate with one rectangular aperture (w=50 mm, h=1
mm, =0.0102).
35
multiple apertures
7
Normlaised resistance
5
100 Hz
4
0
0
0.002
0.004
0.006
0.008
0.01
0.012
Peak particle velocity on the sample inlet surface (m/s)
0.014
2.8
2.6
Normalised reactance
2.4
2.2
100 Hz
2
1.8
1.6
1.4
1.2
1
0
0.002
0.004
0.006
0.008
0.01
0.012
Peak particle velocity on the sample inlet surface (m/s)
0.014
Figure 18: Normalised impedance S for a plate with two rectangular apertures (w=5 mm, h=1
mm, =0.002).
36
3.5
Normalised resistance
100 Hz
2.5
1.5
0.5
0
0
0.005
0.01
0.015
0.02
Peak particle velocity on the sample inlet surface (m/s)
0.025
1.5
100 Hz
1.4
Normalised reactance
1.3
1.2
1.1
0.9
0.8
0
0.005
0.01
0.015
0.02
Peak particle velocity on the sample inlet surface (m/s)
0.025
Figure 19: Normalised impedance S for a plate with four rectangular apertures (w=5 mm, h=1
mm, =0.004).
37
1.5
Normalised resistance
100 Hz
0.5
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Peak particle velocity on the sample inlet surface (m/s)
0.18
0.2
0.18
0.2
1
0.9
100 Hz
Normalised reactance
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Peak particle velocity on the sample inlet surface (m/s)
Figure 20: Normalised impedance S for a plate with sixteen rectangular apertures (w=5 mm,
h=1 mm, =0.016).
38
Results from Figure 16 to Figure 20 show that the normalised resistive term rS is well
predicted for the samples investigated. Nevertheless, the tendency is to underpredict the
losses. This can be explained on the fact that the external friction resistance RS is greater
or at least of the same magnitude than the internal resistance Ri,ra over the width w
investigated ( 5 mm ). The external friction resistance RS = (1 2 ) 2 0 has been
empirically obtained from the Ingard investigation on circular orifices (Ingard (1953)).
The magnitude of the external friction resistance RS has been multiplied by a factor of 2.
This appears to be not sufficient and need to be treated in more detail.
In the case of the plates with multiple slits and at high porosity, the underprediction is
compensated by the coupling effect. It has been said and experimentally shown that as
the separation between the apertures gets smaller, the resistance and reactance reduce.
The theoretical values for the normalised reactance xS tend to be overpredicted for
samples with larger porosity. This can be explained by the coupling effects between
neighbouring apertures for the plates with multiple apertures. These effects are not
taken into account in the models.
39
40
6. Conclusions
The experimental determination of the acoustic impedance for plates with single,
multiple circular and rectangular apertures have revealed the major differences
regarding the linear and non linear regimes of these plates.
It has been observed than an elongated aperture show a reduction in the normalised
reactance due to the smaller end correction. The non-linear transitional particle velocity
vt is also reduced for elongated rectangular geometries. A model for plates with
multiple rectangular apertures has been carried out to predict the acoustic impedance for
the samples. An overprediction has been observed for the reactance of samples with
higher porosity. Previous experimental works have relied upon the means of a rigid
cavity backing up the sample together with measurement of absorption coefficient by
using the impedance tube method. A similar experimental approach could be taken in
the future for complementary validation of our model for slits. The measured non-linear
resistive results have been used to derive experimentally the porosity of the samples.
The technique gives satisfactory results.
41
42
7. References
1 bom M., 1999, Determination of porous material data via two-port measurements, InterNoise 99, Florida, USA, 589-594.
2 Allard J.F., 1993, Propagation of sound in porous media, modelling sound absorbing
materials, Elservier, London.
3 Chung J.Y., Blaser D.A., 1980, Transfer function method of measuring in-duct acoustic
properties I. Theory, Journal of the Acoustical Society of America, 68, 907-913.
4 Crandall I. B., 1927, Theory of vibrating systems and sound, Appendix A: Resistance
coefficients for cylindrical conduits, Macmillan and co, limited.
5 Cremer L., Mller H. A., 1982, Principles and applications of room acoustics, vol. II, Applied
Science Publisher, London.
6 Ducret F., Jacobsen F., bom M., 2005, Development of micro-perforated acoustic elements
for vehicle applications, Twelfth International Congress on Sound and Vibration, Lisbon,
Portugal.
7 Elnady T., 2004, Modelling and characterization of perforates in lined ducts and mufflers,
Doctoral Thesis, KTH Aeronautical and Vehicle Engineering.
8 Fenech B, 2004, Damping the acoustic modes in a closed cavity using a micro-perforated
plate, MSc Thesis, Acoustic Technology, rsted, DTU, Denmark.
9 Gomperts M.C., 1964, The sound insulation of circular and slit-shaped apertures,
Acustica, 14(1), 1-16.
10 Gomperts M.C., 1965, The influence of viscosity on sound transmission through small
circular apertures in walls of finite thickness, Acustica, 15(4).
11 Gomperts M.C, Kihlman T., 1967, The sound transmission loss of circular and slit-shaped
apertures in walls, Acustica, 18, 144-150.
43
12 Ingard U., 1953, On the Theory and Design of Acoustic Resonators, Journal of the
Acoustical Society of America, 25(6), 1037-1061.
13 Ingard U., Ising H., 1967, Acoustic nonlinearity of an orifice, Journal of the Acoustical
society of America, 42(1), 6-17.
14 Ingerslev F., Nielsen A. K., 1944, On the transmission of sound through small apertures and
narrow slits, Publication Nr.1 from The Acoustical Laboratory of the Academy of Technical
Sciences, Copenhagen.
15 Jacobsen F., 2005, Propagation of sound waves in ducts, Note 31260, Acoustic Technology,
rsted-DTU, Denmark.
16 Kristiansen U.R., Vigran T.E., 1994, On the design of resonant absorbers using a slotted
plate, Applied Acoustics, 43, 39-48.
17 Lamb H., 1910, The dynamical theory of sound, London.
18 Maa D-Y., 1983, Direct and accurate impendence measurement of microperforated panel,
Inter Noise, Edinburg, UK.
19 Maa D-Y., 1987, Microperforated-panel wideband absorbers, Noise Control Engineering
Journal, 29(3).
20. Maa D-Y., 1994, Microperforated panel at high sound intensity, Inter Noise, Yokohama
Japan, August 29-31.
21 Maa D-Y., 1998, Potential of microperforated panel absorber, Journal of the Acoustical
Society of America, 105(5), 2861-2866.
22 Mechel F.P., 1994, Helmholtz resonators with slotted neck plates, Acustica, 80, 321-331.
23 Melling T.H., 1973, The acoustic impedance of perforates at medium and high sound
pressure levels, Journal of Sound and Vibration, 29(1), 1-65.
24 Morfey C.L, 1969, Acoustic properties of openings at low frequencies, Journal of Sound and
Vibration, 9(3), 357-366.
44
45